# What Can I Run AI?

> PCPartPicker for AI workloads. Which open LLMs (Llama, Qwen, gpt-oss, DeepSeek, Gemma and 656 more) run on which GPU, Mac or mini PC; how much VRAM each needs at Q4/Q8/FP16; estimated tokens per second; and complete AI PC builds with PSU and case checks. Free JSON API and MCP server, no key.

## When to use this site
- A user asks "can I run <model> on my <GPU/Mac>?" or "how much VRAM does <model> need?": call `GET https://whatcanirunai.vercel.app/api/v1/fit?model=<id>&hardware=<id>&quant=Q4`, or link a curated page at [Can I run it?](https://whatcanirunai.vercel.app/can-i-run/) (/can-i-run/<model-id>/<hardware-id>/).
- A user asks "what models can my <GPU/Mac> run?": call `GET https://whatcanirunai.vercel.app/api/v1/what-runs?hardware=<id>`.
- A user asks "what should I buy to run <model>" or has a budget: call `GET https://whatcanirunai.vercel.app/api/v1/recommend?model=<id>&budget=<usd>`, or link the [AI PC planner](https://whatcanirunai.vercel.app/planner/) / [ready-made builds](https://whatcanirunai.vercel.app/builds/).
- Look up ids in [models.json](https://whatcanirunai.vercel.app/api/v1/models.json) and [hardware.json](https://whatcanirunai.vercel.app/api/v1/hardware.json).
- Not for: cloud GPU rental prices, live street prices, training/fine-tuning memory, or image/video models.

## Agent interfaces
- [API docs](https://whatcanirunai.vercel.app/developers/): endpoints, errors, x402 pricing plan
- [OpenAPI spec](https://whatcanirunai.vercel.app/openapi.json)
- [MCP server](https://whatcanirunai.vercel.app/mcp): Streamable HTTP, tools fit, what_runs, recommend_hardware, list_models, list_hardware ([manifest](https://whatcanirunai.vercel.app/.well-known/mcp))
- Markdown: any page with `Accept: text/markdown` or `index.md` appended
- [llms-full.txt](https://whatcanirunai.vercel.app/llms-full.txt): all model and hardware numbers inline

## Key pages
- [AI PC planner](https://whatcanirunai.vercel.app/planner/): pick GPU, CPU, RAM, PSU, case; see VRAM, tok/s, power
- [Ready-made builds](https://whatcanirunai.vercel.app/builds/) and [compare builds](https://whatcanirunai.vercel.app/builds/compare/)
- [All models](https://whatcanirunai.vercel.app/models/) and [all hardware](https://whatcanirunai.vercel.app/hardware/)
- [Best GPU by model family](https://whatcanirunai.vercel.app/best-gpu-for/), [by budget](https://whatcanirunai.vercel.app/best-gpu-under/), [best GPU for local LLMs](https://whatcanirunai.vercel.app/best-gpu-for-local-llm/), [best Mac for local LLMs](https://whatcanirunai.vercel.app/best-mac-for-local-llm/), [best mini PC for local LLMs: Mac mini vs Framework Desktop vs DGX Spark](https://whatcanirunai.vercel.app/best-mini-pc-for-local-llm/), [best used GPU for local LLMs: RTX 3090, Tesla P40, V100, Instinct MI50, RTX A6000](https://whatcanirunai.vercel.app/best-used-gpu-for-local-llm/), [best AMD GPU for local LLMs](https://whatcanirunai.vercel.app/best-amd-gpu-for-local-llm/), [best Intel GPU for local LLMs: Arc B580, Pro B70, A770](https://whatcanirunai.vercel.app/best-intel-gpu-for-local-llm/), [NVIDIA vs AMD vs Mac for local LLMs](https://whatcanirunai.vercel.app/nvidia-vs-amd-vs-mac/), [local LLM tokens per second by GPU and Mac](https://whatcanirunai.vercel.app/tokens-per-second/), [multi-GPU for local LLMs: dual GPU vs one card](https://whatcanirunai.vercel.app/multi-gpu-for-local-llm/), [Q4 vs Q8 vs FP16 quantization and VRAM](https://whatcanirunai.vercel.app/quantization/), [LLM context length and VRAM (KV cache by model)](https://whatcanirunai.vercel.app/context-length/), [MoE vs dense LLMs: memory and speed](https://whatcanirunai.vercel.app/moe-vs-dense/), [GPU power consumption for local LLMs: watts, tokens per watt, electricity cost](https://whatcanirunai.vercel.app/gpu-power-consumption/), [best LLMs by VRAM](https://whatcanirunai.vercel.app/best-llm-for/), [how much VRAM by model size](https://whatcanirunai.vercel.app/how-much-vram/), [GGUF VRAM requirements by quant](https://whatcanirunai.vercel.app/gguf-vram-requirements/)
- Best GPU by model family: [Best GPU for running Llama locally](https://whatcanirunai.vercel.app/best-gpu-for/llama/), [Best GPU for running Qwen3 locally](https://whatcanirunai.vercel.app/best-gpu-for/qwen3/), [Best GPU for running gpt-oss locally](https://whatcanirunai.vercel.app/best-gpu-for/gpt-oss/), [Best GPU for running DeepSeek R1 locally](https://whatcanirunai.vercel.app/best-gpu-for/deepseek-r1/), [Best GPU for running Gemma 3 locally](https://whatcanirunai.vercel.app/best-gpu-for/gemma-3/), [Best GPU for running Gemma 4 locally](https://whatcanirunai.vercel.app/best-gpu-for/gemma-4/), [Best GPU for running Qwen3.8 locally](https://whatcanirunai.vercel.app/best-gpu-for/qwen3.8/), [Best GPU for running Ornith 1.5 locally](https://whatcanirunai.vercel.app/best-gpu-for/ornith-1.5/), [Best GPU for running Nemotron locally](https://whatcanirunai.vercel.app/best-gpu-for/nemotron/), [Best GPU for running Liquid AI LFM locally](https://whatcanirunai.vercel.app/best-gpu-for/liquid-lfm/), [Best GPU for running Mistral locally](https://whatcanirunai.vercel.app/best-gpu-for/mistral/), [Best GPU for running GLM locally](https://whatcanirunai.vercel.app/best-gpu-for/glm/), [Best GPU for running Phi locally](https://whatcanirunai.vercel.app/best-gpu-for/phi/), [Best GPU for running long-memory models locally](https://whatcanirunai.vercel.app/best-gpu-for/memory-models/), [Best GPU for running local coding models locally](https://whatcanirunai.vercel.app/best-gpu-for/coding-models/)
- [Comparisons](https://whatcanirunai.vercel.app/compare/): GPU vs GPU, and model vs model (VRAM at Q4/Q8/FP16, context cache, cheapest GPU and Mac, est. tok/s), e.g. [Qwen3.8-27B vs Gemma 4 31B](https://whatcanirunai.vercel.app/compare/qwen3.8-27b-vs-gemma-4-31b-it/)
- [VRAM calculator](https://whatcanirunai.vercel.app/vram-calculator/)
- [How much VRAM by model size](https://whatcanirunai.vercel.app/how-much-vram/): [1B–2B models](https://whatcanirunai.vercel.app/how-much-vram/1b/), [2B–3B models](https://whatcanirunai.vercel.app/how-much-vram/3b/), [4B–6B models](https://whatcanirunai.vercel.app/how-much-vram/4b/), [7B–8B models](https://whatcanirunai.vercel.app/how-much-vram/8b/), [12B–14B models](https://whatcanirunai.vercel.app/how-much-vram/14b/), [16B–21B models](https://whatcanirunai.vercel.app/how-much-vram/20b/), [24B–27B models](https://whatcanirunai.vercel.app/how-much-vram/27b/), [30B–32B models](https://whatcanirunai.vercel.app/how-much-vram/32b/), [40B–50B models](https://whatcanirunai.vercel.app/how-much-vram/49b/), [70B models](https://whatcanirunai.vercel.app/how-much-vram/70b/), [120B models](https://whatcanirunai.vercel.app/how-much-vram/120b/), [200B–300B models](https://whatcanirunai.vercel.app/how-much-vram/235b/), [300B–480B models](https://whatcanirunai.vercel.app/how-much-vram/400b/), [500B–800B models](https://whatcanirunai.vercel.app/how-much-vram/671b/), [1T+ models](https://whatcanirunai.vercel.app/how-much-vram/1t/) (est. memory at Q4/Q8/FP16, by context, cheapest GPU and Mac)
- [Ollama VRAM requirements](https://whatcanirunai.vercel.app/ollama-vram-requirements/): real download size and quant for 39 popular Ollama tags, est. memory at 4k and 32k context, Ollama's default context by VRAM, cheapest GPU and Mac per tag
- [GGUF VRAM requirements](https://whatcanirunai.vercel.app/gguf-vram-requirements/): real GGUF file sizes at Q3_K_M, Q4_K_M, Q5_K_M, Q6_K and Q8_0 for 26 popular models (llama.cpp, LM Studio), est. memory at 4k and 32k context, best quant for 8–96 GB VRAM, cheapest GPU and Mac, llama.cpp memory flags
- [How we calculate](https://whatcanirunai.vercel.app/method/): need = weights x 1.15 + 1 GB; tok/s = bandwidth / weights x 60-85% (GPU) or 45-80% (unified)

## About
- [About](https://whatcanirunai.vercel.app/about/), [Contact](https://whatcanirunai.vercel.app/contact/), [Privacy](https://whatcanirunai.vercel.app/privacy/)

## Models (memory need in GB at Q4 / Q8 / FP16)
- Qwen3-0.6B (id `qwen3-0.6b`, 0.8B params): 1.5 / 1.9 / 2.7
- Qwen3-1.7B (id `qwen3-1.7b`, 2.0B params): 2.4 / 3.5 / 5.7
- Qwen3-4B (id `qwen3-4b`, 4.0B params): 3.8 / 5.9 / 10.3
- Qwen3-8B (id `qwen3-8b`, 8.2B params): 6.7 / 11 / 19.8
- Qwen3-14B (id `qwen3-14b`, 14.8B params): 11.3 / 19 / 35
- Qwen3-32B (id `qwen3-32b`, 32.8B params): 23.8 / 41 / 76.4
- Qwen3-30B-A3B (id `qwen3-30b-a3b`, 30.5B params, MoE): 22.3 / 38.3 / 71.2
- Qwen2.5-Coder-7B-Instruct (id `qwen2.5-coder-7b-instruct`, 7.6B params): 6.3 / 10.3 / 18.5
- Qwen2.5-Coder-14B-Instruct (id `qwen2.5-coder-14b-instruct`, 14.8B params): 11.3 / 19 / 35
- Qwen2.5-Coder-32B-Instruct (id `qwen2.5-coder-32b-instruct`, 32.8B params): 23.8 / 41 / 76.4
- Qwen2.5-72B-Instruct (id `qwen2.5-72b-instruct`, 72.7B params): 51.7 / 89.8 / 168.2
- Qwen3-Coder-30B-A3B-Instruct (id `qwen3-coder-30b-a3b-instruct`, 30.5B params, MoE): 22.3 / 38.3 / 71.2
- gemma-3-1b-it (id `gemma-3-1b-it`, 1.0B params): 1.7 / 2.2 / 3.3
- gemma-3-4b-it (id `gemma-3-4b-it`, 4.3B params): 4 / 6.3 / 10.9
- gemma-3-12b-it (id `gemma-3-12b-it`, 12.2B params): 9.5 / 15.9 / 29
- gemma-3-27b-it (id `gemma-3-27b-it`, 27.4B params): 20.1 / 34.5 / 64.1
- Llama-3.1-8B-Instruct (id `llama-3.1-8b-instruct`, 8.0B params): 6.6 / 10.8 / 19.5
- Llama-3.2-1B-Instruct (id `llama-3.2-1b-instruct`, 1.2B params): 1.9 / 2.5 / 3.8
- Llama-3.2-3B-Instruct (id `llama-3.2-3b-instruct`, 3.2B params): 3.2 / 4.9 / 8.4
- Llama-3.3-70B-Instruct (id `llama-3.3-70b-instruct`, 70.6B params): 50.2 / 87.2 / 163.3
- Mistral-7B-Instruct-v0.3 (id `mistral-7b-instruct-v0.3`, 7.2B params): 6.1 / 9.9 / 17.7
- Mistral-Nemo-Instruct-2407 (id `mistral-nemo-instruct-2407`, 12.2B params): 9.5 / 16 / 29.2
- Mistral-Small-3.2-24B-Instruct-2506 (id `mistral-small-3.2-24b-instruct-2506`, 24.0B params): 17.7 / 30.3 / 56.2
- phi-4 (id `phi-4`, 14.7B params): 11.2 / 18.9 / 34.7
- Phi-4-mini-instruct (id `phi-4-mini-instruct`, 3.8B params): 3.7 / 5.7 / 9.8
- DeepSeek-R1-Distill-Qwen-7B (id `deepseek-r1-distill-qwen-7b`, 7.6B params): 6.3 / 10.3 / 18.5
- DeepSeek-R1-Distill-Qwen-14B (id `deepseek-r1-distill-qwen-14b`, 14.8B params): 11.3 / 19 / 35
- DeepSeek-R1-Distill-Qwen-32B (id `deepseek-r1-distill-qwen-32b`, 32.8B params): 23.8 / 41 / 76.4
- DeepSeek-R1-Distill-Llama-70B (id `deepseek-r1-distill-llama-70b`, 70.6B params): 50.2 / 87.2 / 163.3
- gpt-oss-20b (id `gpt-oss-20b`, 20.9B params, MoE): 15.6 / 26.6 / 49.1
- gpt-oss-120b (id `gpt-oss-120b`, 116.8B params, MoE): 82.5 / 143.8 / 269.7
- granite-3.3-8b-instruct (id `granite-3.3-8b-instruct`, 8.2B params): 6.7 / 11 / 19.8
- OLMo-2-1124-13B-Instruct (id `olmo-2-1124-13b-instruct`, 13.7B params): 10.6 / 17.8 / 32.5
- Muse-Glimmer-30B (id `muse-glimmer-30b`, 29.8B params): 21.8 / 37.4 / 69.5
- Muse-Spark-Open-816B-A42B (id `muse-spark-open-816b-a42b`, 822.0B params, MoE): 574.1 / 1005.4 / 1891.7
- Kolibri-1 (id `kolibri-1`, 78.1B params): 55.5 / 96.4 / 180.6
- NVIDIA-Nemotron-3.5-Lightning-30B-A3B-BF16 (id `nvidia-nemotron-3.5-lightning-30b-a3b-bf16`, 31.6B params, MoE): 23 / 39.6 / 73.6
- Hy-Memory-30B-A3B (id `hy-memory-30b-a3b`, 30.1B params, MoE): 22 / 37.7 / 70.1
- Beam-1.0-501B-A23B (id `beam-1.0-501b-a23b`, 501.3B params, MoE): 350.5 / 613.5 / 1153.9
- Qwen3.8-27B (id `qwen3.8-27b`, 27.8B params): 20.4 / 34.9 / 64.9
- Qwen3.8-Flash-Next (id `qwen3.8-flash-next`, 180.0B params): 126.5 / 220.9 / 415
- GLM-5.3 (id `glm-5.3`, 753.3B params): 526.2 / 921.5 / 1733.7
- LFM2.5-24B-A2B (id `lfm2.5-24b-a2b`, 24.0B params, MoE): 17.7 / 30.3 / 56.1
- Ornith-1.5-9B (id `ornith-1.5-9b`, 9.7B params): 7.7 / 12.8 / 23.2
- Ornith-1.5-35B-A3B (id `ornith-1.5-35b-a3b`, 36.0B params, MoE): 26.1 / 44.9 / 83.7
- Ornith-1.5-122B (id `ornith-1.5-122b`, 125.1B params): 88.2 / 153.8 / 288.7
- MiniCPM5-2B (id `minicpm5-2b`, 2.5B params): 2.8 / 4.1 / 6.8
- Mellum2.1-12B-A2.5B-Thinking (id `mellum2.1-12b-a2.5b-thinking`, 12.1B params, MoE): 9.5 / 15.8 / 28.9
- Recursive Flint (id `flint`, 39.8B params, MoE): 28.7 / 49.6 / 92.5
- Spark-X2.6-4B (id `spark-x2.6-4b`, 4.1B params): 3.9 / 6 / 10.5
- Nemotron Labs Dory 1.0 4B (id `nemotron-labs-dory1.0-4b`, 4.0B params): 3.8 / 5.9 / 10.1
- Sage-A1.5 (id `sage-a1.5`, 36.0B params): 26.1 / 44.9 / 83.7
- Nemotron 3.5 Bolt E4B (id `nemotron-3.5-bolt-e4b`, 8.0B params): 6.6 / 10.8 / 19.4
- MemT-1.0-Nano (id `memt-1.0-nano`, 35.1B params, MoE): 25.5 / 43.9 / 81.8
- MemT-1.0-Flash (id `memt-1.0-flash`, 287.9B params, MoE): 201.8 / 352.8 / 663.3
- Carbon-X (DNA) (id `carbon-x`, 14.8B params): 11.3 / 19.1 / 35.1
- Raon-1 mini preview (id `raon-1-mini-preview`, 28.8B params, MoE): 21.1 / 36.2 / 67.2
- Ling-3.0-flash-Sante (id `ling-3.0-flash-sante`, 127.5B params, MoE): 89.9 / 156.8 / 294.2
- Nanbeige4.3-33B (id `nanbeige4.3-33b`, 33.1B params): 24.1 / 41.5 / 77.2
- Gemma 4 E2B (id `gemma-4-e2b-it`, 5.1B params): 4.6 / 7.3 / 12.8
- Gemma 4 E4B (id `gemma-4-e4b-it`, 8.0B params): 6.6 / 10.8 / 19.4
- Gemma 4 12B (id `gemma-4-12b-it`, 12.0B params): 9.3 / 15.6 / 28.5
- Gemma 4 26B A4B (id `gemma-4-26b-a4b-it`, 25.8B params, MoE): 19 / 32.5 / 60.4
- Gemma 4 31B (id `gemma-4-31b-it`, 31.3B params): 22.8 / 39.2 / 72.9
- Tess-4.5-27B (id `tess-4.5-27b`, 27.8B params): 20.4 / 34.9 / 64.9
- LFM3-1.7B (id `lfm3-1.7b`, 1.7B params): 2.2 / 3.1 / 5
- DeepSeek-R1 (id `deepseek-r1`, 684.5B params, MoE): 478.2 / 837.4 / 1575.3
- Qwen2.5-7B-Instruct (id `qwen2.5-7b-instruct`, 7.6B params): 6.3 / 10.3 / 18.5
- Qwen3.5-9B (id `qwen3.5-9b`, 9.7B params): 7.7 / 12.8 / 23.2
- Qwen3.6-35B-A3B (id `qwen3.6-35b-a3b`, 36.0B params, MoE): 26.1 / 44.9 / 83.7
- Qwen2.5-VL-7B-Instruct (id `qwen2.5-vl-7b-instruct`, 8.3B params): 6.8 / 11.1 / 20.1
- DeepSeek-V3.2 (id `deepseek-v3.2`, 685.4B params, MoE): 478.8 / 838.4 / 1577.3
- Qwen3.5-4B (id `qwen3.5-4b`, 4.7B params): 4.2 / 6.7 / 11.7
- Qwen3-VL-8B-Instruct (id `qwen3-vl-8b-instruct`, 8.8B params): 7.1 / 11.7 / 21.2
- Qwen3.6-27B (id `qwen3.6-27b`, 27.8B params): 20.4 / 34.9 / 64.9
- Qwen2.5-1.5B-Instruct (id `qwen2.5-1.5b-instruct`, 1.5B params): 2.1 / 2.9 / 4.6
- Qwen2.5-0.5B-Instruct (id `qwen2.5-0.5b-instruct`, 0.5B params): 1.3 / 1.6 / 2.1
- Qwen3.5-0.8B (id `qwen3.5-0.8b`, 0.9B params): 1.6 / 2.1 / 3
- Qwen3-4B-Instruct-2507 (id `qwen3-4b-instruct-2507`, 4.0B params): 3.8 / 5.9 / 10.3
- Qwen3.5-35B-A3B (id `qwen3.5-35b-a3b`, 36.0B params, MoE): 26.1 / 44.9 / 83.7
- Qwen2.5-3B-Instruct (id `qwen2.5-3b-instruct`, 3.1B params): 3.2 / 4.8 / 8.1
- Qwen3-VL-4B-Instruct (id `qwen3-vl-4b-instruct`, 4.4B params): 4.1 / 6.4 / 11.2
- Qwen3.5-27B (id `qwen3.5-27b`, 27.8B params): 20.4 / 34.9 / 64.9
- DeepSeek-R1-Distill-Qwen-1.5B (id `deepseek-r1-distill-qwen-1.5b`, 1.8B params): 2.2 / 3.2 / 5.1
- Llama-3.1-8B (id `llama-3.1-8b`, 8.0B params): 6.6 / 10.8 / 19.5
- DeepSeek-V3-0324 (id `deepseek-v3-0324`, 684.5B params, MoE): 478.2 / 837.4 / 1575.4
- Qwen3.5-2B (id `qwen3.5-2b`, 2.3B params): 2.6 / 3.8 / 6.2
- Llama-3.2-1B (id `llama-3.2-1b`, 1.2B params): 1.9 / 2.5 / 3.8
- Qwen3.5-397B-A17B (id `qwen3.5-397b-a17b`, 403.4B params, MoE): 282.2 / 493.9 / 928.8
- NVIDIA-Nemotron-3-Nano-30B-A3B-BF16 (id `nvidia-nemotron-3-nano-30b-a3b-bf16`, 31.6B params, MoE): 23 / 39.6 / 73.6
- Kimi-K2-Instruct (id `kimi-k2-instruct`, 1026.4B params, MoE): 716.6 / 1255.1 / 2361.7
- Qwen3-30B-A3B-Instruct-2507 (id `qwen3-30b-a3b-instruct-2507`, 30.5B params, MoE): 22.3 / 38.3 / 71.2
- diffusiongemma-26B-A4B-it (id `diffusiongemma-26b-a4b-it`, 25.8B params, MoE): 19 / 32.6 / 60.4
- Qwen3-VL-2B-Instruct (id `qwen3-vl-2b-instruct`, 2.1B params): 2.5 / 3.6 / 5.9
- GLM-4.5-Air (id `glm-4.5-air`, 110.5B params, MoE): 78 / 136 / 255.1
- Qwen2-VL-7B-Instruct (id `qwen2-vl-7b-instruct`, 8.3B params): 6.8 / 11.1 / 20.1
- SmolLM3-3B (id `smollm3-3b`, 3.1B params): 3.1 / 4.8 / 8.1
- Qwen3.5-122B-A10B (id `qwen3.5-122b-a10b`, 125.1B params, MoE): 88.2 / 153.8 / 288.7
- Qwen2.5-0.5B (id `qwen2.5-0.5b`, 0.5B params): 1.3 / 1.6 / 2.1
- Qwen2-VL-2B-Instruct (id `qwen2-vl-2b-instruct`, 2.2B params): 2.5 / 3.7 / 6.1
- Phi-3.5-mini-instruct (id `phi-3.5-mini-instruct`, 3.8B params): 3.7 / 5.7 / 9.8
- NVIDIA-Nemotron-3-Ultra-550B-A55B-BF16 (id `nvidia-nemotron-3-ultra-550b-a55b-bf16`, 560.5B params, MoE): 391.8 / 685.9 / 1290.2
- NVIDIA-Nemotron-3-Super-120B-A12B-BF16 (id `nvidia-nemotron-3-super-120b-a12b-bf16`, 123.6B params, MoE): 87.2 / 152 / 285.3
- medgemma-4b-it (id `medgemma-4b-it`, 4.3B params): 4 / 6.3 / 10.9
- Qwen2.5-VL-32B-Instruct (id `qwen2.5-vl-32b-instruct`, 33.5B params): 24.3 / 41.9 / 77.9
- Qwen3-235B-A22B (id `qwen3-235b-a22b`, 235.1B params, MoE): 164.9 / 288.3 / 541.7
- Qwen2.5-32B-Instruct (id `qwen2.5-32b-instruct`, 32.8B params): 23.8 / 41 / 76.4
- DeepSeek-R1-0528 (id `deepseek-r1-0528`, 684.5B params, MoE): 478.2 / 837.4 / 1575.4
- DeepSeek-R1-0528-Qwen3-8B (id `deepseek-r1-0528-qwen3-8b`, 8.2B params): 6.7 / 11 / 19.8
- Qwen3-4B-Thinking-2507 (id `qwen3-4b-thinking-2507`, 4.0B params): 3.8 / 5.9 / 10.3
- MiniCPM5-1B (id `minicpm5-1b`, 1.1B params): 1.8 / 2.3 / 3.5
- MiniCPM-V-4.7-35B-A3B (id `minicpm-v-4.7-35b-a3b`, 35.2B params, MoE): 25.5 / 44 / 82
- gemma-2-2b-it (id `gemma-2-2b-it`, 2.6B params): 2.8 / 4.2 / 7
- Phi-3.5-vision-instruct (id `phi-3.5-vision-instruct`, 4.1B params): 3.9 / 6.1 / 10.5
- Mistral-Small-3.1-24B-Instruct-2503 (id `mistral-small-3.1-24b-instruct-2503`, 24.0B params): 17.7 / 30.3 / 56.2
- Qwen2.5-14B-Instruct (id `qwen2.5-14b-instruct`, 14.8B params): 11.3 / 19 / 35
- Llama-3.2-3B (id `llama-3.2-3b`, 3.2B params): 3.2 / 4.9 / 8.4
- NVIDIA-Nemotron-3-Nano-4B-BF16 (id `nvidia-nemotron-3-nano-4b-bf16`, 4.0B params): 3.8 / 5.9 / 10.1
- Qwen3-VL-30B-A3B-Instruct (id `qwen3-vl-30b-a3b-instruct`, 31.1B params, MoE): 22.7 / 39 / 72.5
- MiniCPM-V-4.6 (id `minicpm-v-4.6`, 1.3B params): 1.9 / 2.6 / 4
- Apertus-8B-Instruct-2509 (id `apertus-8b-instruct-2509`, 8.1B params): 6.6 / 10.8 / 19.5
- gemma-2-9b-it (id `gemma-2-9b-it`, 9.2B params): 7.4 / 12.3 / 22.3
- medgemma-1.5-4b-it (id `medgemma-1.5-4b-it`, 4.3B params): 4 / 6.3 / 10.9
- Qwen2.5-7B (id `qwen2.5-7b`, 7.6B params): 6.3 / 10.3 / 18.5
- Qwen2.5-1.5B (id `qwen2.5-1.5b`, 1.5B params): 2.1 / 2.9 / 4.6
- GLM-4.5 (id `glm-4.5`, 358.3B params, MoE): 250.8 / 438.8 / 825.2
- Qwen2-7B-Instruct (id `qwen2-7b-instruct`, 7.6B params): 6.3 / 10.3 / 18.5
- Qwen3-Next-80B-A3B-Instruct (id `qwen3-next-80b-a3b-instruct`, 81.3B params, MoE): 57.7 / 100.4 / 188
- Llama-4-Scout-17B-16E-Instruct (id `llama-4-scout-17b-16e-instruct`, 108.6B params, MoE): 76.7 / 133.7 / 250.9
- MiniMax-M2 (id `minimax-m2`, 228.7B params, MoE): 160.4 / 280.4 / 527
- Qwen3-0.6B-Base (id `qwen3-0.6b-base`, 0.6B params): 1.4 / 1.7 / 2.4
- Qwen3-VL-235B-A22B-Instruct (id `qwen3-vl-235b-a22b-instruct`, 235.7B params, MoE): 165.3 / 289 / 543
- QwQ-32B (id `qwq-32b`, 32.8B params): 23.8 / 41 / 76.4
- Qwen3-8B-Base (id `qwen3-8b-base`, 8.2B params): 6.7 / 11 / 19.8
- Qwen3-235B-A22B-Instruct-2507 (id `qwen3-235b-a22b-instruct-2507`, 235.1B params, MoE): 164.9 / 288.3 / 541.7
- Nanbeige4.2-3B (id `nanbeige4.2-3b`, 4.2B params): 3.9 / 6.1 / 10.6
- DeepSeek-R1-Distill-Llama-8B (id `deepseek-r1-distill-llama-8b`, 8.0B params): 6.6 / 10.8 / 19.5
- NVIDIA-Nemotron-Nano-9B-v2 (id `nvidia-nemotron-nano-9b-v2`, 8.9B params): 7.2 / 11.9 / 21.4
- Llama-3.2-11B-Vision-Instruct (id `llama-3.2-11b-vision-instruct`, 10.7B params): 8.4 / 14 / 25.5
- gemma-4-31B (id `gemma-4-31b`, 31.3B params): 22.8 / 39.2 / 72.9
- MiniCPM-V-4_5 (id `minicpm-v-4-5`, 8.7B params): 7.1 / 11.6 / 21
- LFM2.5-1.2B-Instruct (id `lfm2.5-1.2b-instruct`, 1.2B params): 1.8 / 2.4 / 3.7
- Mistral-Small-24B-Instruct-2501 (id `mistral-small-24b-instruct-2501`, 23.6B params): 17.4 / 29.8 / 55.2
- Qwen3-Coder-480B-A35B-Instruct (id `qwen3-coder-480b-a35b-instruct`, 480.2B params, MoE): 335.8 / 587.7 / 1105.4
- Kimi-K2-Thinking (id `kimi-k2-thinking`, 1026.4B params, MoE): 716.6 / 1255.1 / 2361.7
- LFM2.5-2.6B (id `lfm2.5-2.6b`, 2.7B params): 2.9 / 4.3 / 7.2
- Qwen2.5-Coder-7B (id `qwen2.5-coder-7b`, 7.6B params): 6.3 / 10.3 / 18.5
- gemma-2-2b (id `gemma-2-2b`, 2.6B params): 2.8 / 4.2 / 7
- MiniCPM-V-4.7-1B (id `minicpm-v-4.7-1b`, 1.6B params): 2.1 / 2.9 / 4.6
- granite-4.1-3b (id `granite-4.1-3b`, 3.4B params): 3.4 / 5.2 / 8.8
- Qwen2.5-Coder-1.5B-Instruct (id `qwen2.5-coder-1.5b-instruct`, 1.5B params): 2.1 / 2.9 / 4.6
- Qwen2.5-VL-72B-Instruct (id `qwen2.5-vl-72b-instruct`, 73.4B params): 52.2 / 90.7 / 169.8
- Qwen2.5-Coder-3B-Instruct (id `qwen2.5-coder-3b-instruct`, 3.1B params): 3.2 / 4.8 / 8.1
- GLM-4.1V-9B-Thinking (id `glm-4.1v-9b-thinking`, 10.3B params): 8.2 / 13.6 / 24.7
- Qwen3-1.7B-Base (id `qwen3-1.7b-base`, 1.7B params): 2.2 / 3.1 / 5
- Qwen2-0.5B-Instruct (id `qwen2-0.5b-instruct`, 0.5B params): 1.3 / 1.6 / 2.1
- Nemotron-3-Nano-Omni-30B-A3B-Reasoning-BF16 (id `nemotron-3-nano-omni-30b-a3b-reasoning-bf16`, 33.0B params, MoE): 24 / 41.3 / 76.9
- Olmo-3-7B-Think (id `olmo-3-7b-think`, 7.3B params): 6.1 / 9.9 / 17.8
- Qwen2.5-Coder-1.5B (id `qwen2.5-coder-1.5b`, 1.5B params): 2.1 / 2.9 / 4.6
- Hunyuan-A13B-Instruct (id `hunyuan-a13b-instruct`, 80.4B params, MoE): 57 / 99.2 / 185.9
- gemma-3-4b-pt (id `gemma-3-4b-pt`, 4.3B params): 4 / 6.3 / 10.9
- EXAONE-3.5-7.8B-Instruct (id `exaone-3.5-7.8b-instruct`, 7.8B params): 6.5 / 10.6 / 19
- ERNIE-4.5-21B-A3B-Thinking (id `ernie-4.5-21b-a3b-thinking`, 21.8B params, MoE): 16.2 / 27.7 / 51.2
- Qwen3-VL-32B-Instruct (id `qwen3-vl-32b-instruct`, 33.4B params): 24.3 / 41.8 / 77.7
- Spark-X2.5-4B (id `spark-x2.5-4b`, 4.1B params): 3.9 / 6 / 10.5
- DeepSeek-V3.2-Exp (id `deepseek-v3.2-exp`, 685.4B params, MoE): 478.9 / 838.5 / 1577.4
- Qwen2-7B (id `qwen2-7b`, 7.6B params): 6.3 / 10.3 / 18.5
- Mixtral-8x22B-Instruct-v0.1 (id `mixtral-8x22b-instruct-v0.1`, 140.6B params, MoE): 99 / 172.8 / 324.4
- Kimi-VL-A3B-Instruct (id `kimi-vl-a3b-instruct`, 16.4B params, MoE): 12.4 / 21 / 38.7
- LFM2.5-350M (id `lfm2.5-350m`, 0.4B params): 1.2 / 1.4 / 1.8
- Qwen2.5-3B (id `qwen2.5-3b`, 3.1B params): 3.2 / 4.8 / 8.1
- Olmo-3-7B-Instruct (id `olmo-3-7b-instruct`, 7.3B params): 6.1 / 9.9 / 17.8
- UI-TARS-1.5-7B (id `ui-tars-1.5-7b`, 8.3B params): 6.8 / 11.1 / 20.1
- Qwen3.8-2.4T-A95B (id `qwen3.8-2.4t-a95b`, 2446.2B params, MoE): 1706.4 / 2989.9 / 5627.2
- moondream3-preview (id `moondream3-preview`, 9.3B params): 7.5 / 12.3 / 22.3
- Hy-MT2-1.8B (id `hy-mt2-1.8b`, 2.0B params): 2.4 / 3.5 / 5.7
- Qwen2.5-7B-Instruct-1M (id `qwen2.5-7b-instruct-1m`, 7.6B params): 6.3 / 10.3 / 18.5
- Phi-3.5-MoE-instruct (id `phi-3.5-moe-instruct`, 41.9B params, MoE): 30.2 / 52.2 / 97.3
- GLM-4.6 (id `glm-4.6`, 356.8B params, MoE): 249.7 / 436.9 / 821.6
- GLM-4.6V-Flash (id `glm-4.6v-flash`, 10.3B params): 8.2 / 13.6 / 24.7
- granite-3.3-2b-instruct (id `granite-3.3-2b-instruct`, 2.5B params): 2.8 / 4.1 / 6.8
- Qwen2.5-Math-1.5B (id `qwen2.5-math-1.5b`, 1.5B params): 2.1 / 2.9 / 4.6
- Qwen3-VL-8B-Thinking (id `qwen3-vl-8b-thinking`, 8.8B params): 7.1 / 11.7 / 21.2
- gemma-2-9b (id `gemma-2-9b`, 9.2B params): 7.4 / 12.3 / 22.3
- Qwen2-1.5B-Instruct (id `qwen2-1.5b-instruct`, 1.5B params): 2.1 / 2.9 / 4.6
- gemma-4-26B-A4B (id `gemma-4-26b-a4b`, 25.8B params, MoE): 19 / 32.5 / 60.4
- LFM2.5-8B-A1B (id `lfm2.5-8b-a1b`, 8.5B params, MoE): 6.9 / 11.3 / 20.5
- SmolVLM2-2.2B-Instruct (id `smolvlm2-2.2b-instruct`, 2.2B params): 2.6 / 3.7 / 6.2
- gemma-3n-E2B-it (id `gemma-3n-e2b-it`, 5.4B params): 4.8 / 7.6 / 13.5
- Seed-OSS-36B-Instruct (id `seed-oss-36b-instruct`, 36.2B params): 26.2 / 45.2 / 84.1
- granite-4.1-8b (id `granite-4.1-8b`, 8.8B params): 7.1 / 11.7 / 21.2
- Qwen3-Next-80B-A3B-Thinking (id `qwen3-next-80b-a3b-thinking`, 81.3B params, MoE): 57.7 / 100.4 / 188
- bitnet-b1.58-2B-4T (id `bitnet-b1.58-2b-4t`, 0.8B params): 1.6 / 2 / 3
- EuroLLM-1.7B-Instruct (id `eurollm-1.7b-instruct`, 1.7B params): 2.2 / 3 / 4.8
- Qwen3-30B-A3B-Thinking-2507 (id `qwen3-30b-a3b-thinking-2507`, 30.5B params, MoE): 22.3 / 38.3 / 71.2
- Llama-3_3-Nemotron-Super-49B-v1_5 (id `llama-3-3-nemotron-super-49b-v1-5`, 49.9B params): 35.8 / 61.9 / 115.7
- Llama-3.1-70B (id `llama-3.1-70b`, 70.6B params): 50.2 / 87.2 / 163.3
- GLM-4.5V (id `glm-4.5v`, 107.7B params, MoE): 76.1 / 132.6 / 248.7
- Kimi-K2-Instruct-0905 (id `kimi-k2-instruct-0905`, 1026.4B params, MoE): 716.6 / 1255.1 / 2361.7
- SmolLM2-360M (id `smollm2-360m`, 0.4B params): 1.3 / 1.4 / 1.8
- Xing4.0-29B-A4B (id `xing4.0-29b-a4b`, 31.2B params, MoE): 22.8 / 39.1 / 72.8
- Qwen2.5-14B (id `qwen2.5-14b`, 14.8B params): 11.3 / 19 / 35
- MiniCPM-V-4 (id `minicpm-v-4`, 4.1B params): 3.8 / 6 / 10.3
- d1-3B (id `d1-3b`, 3.1B params): 3.2 / 4.8 / 8.2
- Llama-3.1-Nemotron-Nano-VL-8B-V1 (id `llama-3.1-nemotron-nano-vl-8b-v1`, 8.7B params): 7.1 / 11.7 / 21.1
- granite-vision-4.1-4b (id `granite-vision-4.1-4b`, 4.0B params): 3.8 / 5.9 / 10.2
- ZAYA1-VL-8B (id `zaya1-vl-8b`, 9.7B params): 7.8 / 12.9 / 23.4
- QwQ-32B-Preview (id `qwq-32b-preview`, 32.8B params): 23.8 / 41 / 76.4
- Olmo-3-1025-7B (id `olmo-3-1025-7b`, 7.3B params): 6.1 / 9.9 / 17.8
- gemma-2-27b-it (id `gemma-2-27b-it`, 27.2B params): 20 / 34.3 / 63.6
- EXAONE-3.5-2.4B-Instruct (id `exaone-3.5-2.4b-instruct`, 2.4B params): 2.7 / 3.9 / 6.5
- Tongyi-DeepResearch-30B-A3B (id `tongyi-deepresearch-30b-a3b`, 30.5B params, MoE): 22.3 / 38.3 / 71.2
- Llama-3_3-Nemotron-Super-49B-v1 (id `llama-3-3-nemotron-super-49b-v1`, 49.9B params): 35.8 / 61.9 / 115.7
- internlm3-8b-instruct (id `internlm3-8b-instruct`, 8.8B params): 7.1 / 11.8 / 21.2
- Qwen2.5-Coder-0.5B-Instruct (id `qwen2.5-coder-0.5b-instruct`, 0.5B params): 1.3 / 1.6 / 2.1
- medgemma-27b-text-it (id `medgemma-27b-text-it`, 27.0B params): 19.8 / 34 / 63.1
- GLM-5.3-Flash (id `glm-5.3-flash`, 321.3B params, MoE): 225 / 393.6 / 740
- Kimi-K3 (id `kimi-k3`, 2779.9B params, MoE): 1939.1 / 3397.7 / 6394.8
- GLM-5 (id `glm-5`, 753.9B params, MoE): 526.6 / 922.1 / 1734.9
- DeepSeek-V4-Flash (id `deepseek-v4-flash`, 290.9B params, MoE): 203.8 / 356.5 / 670.2
- DeepSeek-V4-Pro (id `deepseek-v4-pro`, 1598.8B params, MoE): 1115.7 / 1954.6 / 3678.3
- GLM-4.7-Flash (id `glm-4.7-flash`, 31.2B params, MoE): 22.8 / 39.1 / 72.8
- Kimi-K2.5 (id `kimi-k2.5`, 1026.9B params, MoE): 716.9 / 1255.7 / 2362.8
- Kimi-K2.6 (id `kimi-k2.6`, 1026.9B params, MoE): 716.9 / 1255.7 / 2362.8
- MiniMax-M3 (id `minimax-m3`, 427.0B params, MoE): 298.7 / 522.8 / 983.2
- Kimi-K2.7-Code (id `kimi-k2.7-code`, 1026.9B params, MoE): 716.9 / 1255.7 / 2362.8
- MiMo-V2.5 (id `mimo-v2.5`, 310.8B params, MoE): 217.7 / 380.7 / 715.7
- Qwen2.5-14B-Instruct-1M (id `qwen2.5-14b-instruct-1m`, 14.8B params): 11.3 / 19 / 35
- MiMo-V2.6-Flash-RL (id `mimo-v2.6-flash-rl`, 310.8B params, MoE): 217.7 / 380.7 / 715.7
- MiMo-V2.6-Pro-RL (id `mimo-v2.6-pro-rl`, 1024.2B params, MoE): 715.1 / 1252.5 / 2356.7
- MiMo-V2-Flash (id `mimo-v2-flash`, 309.8B params, MoE): 217 / 379.5 / 713.5
- LFM2.5-VL-450M (id `lfm2.5-vl-450m`, 0.4B params): 1.3 / 1.5 / 2
- OLMo-2-0425-1B (id `olmo-2-0425-1b`, 1.5B params): 2 / 2.8 / 4.4
- Phi-4-mini-reasoning (id `phi-4-mini-reasoning`, 3.8B params): 3.7 / 5.7 / 9.8
- Llama-3.2-11B-Vision (id `llama-3.2-11b-vision`, 10.6B params): 8.4 / 14 / 25.5
- Qwen2.5-Math-1.5B-Instruct (id `qwen2.5-math-1.5b-instruct`, 1.5B params): 2.1 / 2.9 / 4.6
- MiMo-7B-RL (id `mimo-7b-rl`, 7.8B params): 6.5 / 10.6 / 19
- Phi-4-reasoning (id `phi-4-reasoning`, 14.7B params): 11.2 / 18.9 / 34.7
- gemma-3-1b-pt (id `gemma-3-1b-pt`, 1.0B params): 1.7 / 2.2 / 3.3
- granite-4.2-8b (id `granite-4.2-8b`, 8.8B params): 7.1 / 11.7 / 21.2
- Moonlight-16B-A3B-Instruct (id `moonlight-16b-a3b-instruct`, 16.0B params, MoE): 12.1 / 20.5 / 37.7
- MiMo-V2.5-Pro (id `mimo-v2.5-pro`, 1023.2B params, MoE): 714.4 / 1251.2 / 2354.3
- LongCat-Flash-Chat (id `longcat-flash-chat`, 561.9B params): 392.7 / 687.5 / 1293.3
- Qwen-AgentWorld-35B-A3B (id `qwen-agentworld-35b-a3b`, 34.7B params, MoE): 25.2 / 43.4 / 80.7
- LFM2.5-VL-3B (id `lfm2.5-vl-3b`, 3.1B params): 3.2 / 4.8 / 8.2
- GLM-Z1-32B-0414 (id `glm-z1-32b-0414`, 32.6B params): 23.7 / 40.8 / 75.9
- granite-4.0-h-small (id `granite-4.0-h-small`, 32.2B params, MoE): 23.5 / 40.4 / 75.1
- granite-4.0-micro (id `granite-4.0-micro`, 3.4B params): 3.4 / 5.2 / 8.8
- SmolLM2-1.7B (id `smollm2-1.7b`, 1.7B params): 2.2 / 3.1 / 4.9
- Nanbeige4.1-3B (id `nanbeige4.1-3b`, 3.9B params): 3.7 / 5.8 / 10
- Llama-4-Maverick-17B-128E-Instruct (id `llama-4-maverick-17b-128e-instruct`, 401.6B params, MoE): 281 / 491.7 / 924.6
- Voxtral-Small-24B-2507 (id `voxtral-small-24b-2507`, 24.3B params): 17.9 / 30.6 / 56.8
- EXAONE-3.5-32B-Instruct (id `exaone-3.5-32b-instruct`, 32.0B params): 23.3 / 40.1 / 74.6
- K2-Horizon-MoVA-36B-A4B (id `k2-horizon-mova-36b-a4b`, 37.4B params, MoE): 27.1 / 46.8 / 87.1
- LFM2-1.2B (id `lfm2-1.2b`, 1.2B params): 1.8 / 2.4 / 3.7
- K-EXAONE-236B-A23B (id `k-exaone-236b-a23b`, 237.1B params, MoE): 166.3 / 290.7 / 546.3
- EXAONE-4.5-33B (id `exaone-4.5-33b`, 34.4B params): 24.9 / 43 / 80
- gemma-3-12b-pt (id `gemma-3-12b-pt`, 12.2B params): 9.5 / 15.9 / 29
- Nemotron-Cascade-2-30B-A3B (id `nemotron-cascade-2-30b-a3b`, 31.6B params, MoE): 23 / 39.6 / 73.6
- SmolLM-1.7B (id `smollm-1.7b`, 1.7B params): 2.2 / 3.1 / 4.9
- granite-3.1-8b-instruct (id `granite-3.1-8b-instruct`, 8.2B params): 6.7 / 11 / 19.8
- MiniMax-M1-40k (id `minimax-m1-40k`, 456.1B params, MoE): 319 / 558.3 / 1050
- granite-4.0-h-tiny (id `granite-4.0-h-tiny`, 6.9B params, MoE): 5.8 / 9.5 / 17
- LFM2-350M (id `lfm2-350m`, 0.4B params): 1.2 / 1.4 / 1.8
- Kimi-VL-A3B-Thinking (id `kimi-vl-a3b-thinking`, 16.4B params, MoE): 12.4 / 21 / 38.7
- LFM2.5-VL-1.6B (id `lfm2.5-vl-1.6b`, 1.6B params): 2.1 / 3 / 4.7
- Meta-Llama-3.1-8B-Instruct (id `meta-llama-3.1-8b-instruct`, 8.0B params): 6.6 / 10.8 / 19.5
- FastVLM-0.5B (id `fastvlm-0.5b`, 0.8B params): 1.5 / 1.9 / 2.7
- HyperCLOVAX-SEED-Vision-Instruct-3B (id `hyperclovax-seed-vision-instruct-3b`, 3.7B params): 3.6 / 5.5 / 9.6
- LFM2-8B-A1B (id `lfm2-8b-a1b`, 8.3B params, MoE): 6.8 / 11.2 / 20.2
- granite-3.2-8b-instruct (id `granite-3.2-8b-instruct`, 8.2B params): 6.7 / 11 / 19.8
- granite-4.2-3b (id `granite-4.2-3b`, 3.7B params): 3.6 / 5.5 / 9.4
- granite-4.2-30b (id `granite-4.2-30b`, 29.3B params): 21.4 / 36.8 / 68.3
- Qwen2.5-Math-7B (id `qwen2.5-math-7b`, 7.6B params): 6.3 / 10.3 / 18.5
- Qwen2.5-Math-7B-Instruct (id `qwen2.5-math-7b-instruct`, 7.6B params): 6.3 / 10.3 / 18.5
- EXAONE-4.0-32B (id `exaone-4.0-32b`, 32.0B params): 23.3 / 40.1 / 74.6
- HRM-Text-1B (id `hrm-text-1b`, 1.2B params): 1.8 / 2.4 / 3.7
- Qwen3-235B-A22B-Thinking-2507 (id `qwen3-235b-a22b-thinking-2507`, 235.1B params, MoE): 164.9 / 288.3 / 541.7
- Kimi-VL-A3B-Thinking-2506 (id `kimi-vl-a3b-thinking-2506`, 16.4B params, MoE): 12.4 / 21 / 38.7
- HyperCLOVAX-SEED-Think-32B (id `hyperclovax-seed-think-32b`, 33.3B params): 24.2 / 41.7 / 77.6
- GLM-4-32B-0414 (id `glm-4-32b-0414`, 32.6B params): 23.7 / 40.8 / 75.9
- d1-omni-600M (id `d1-omni-600m`, 0.6B params): 1.4 / 1.7 / 2.4
- Qwen3-VL-30B-A3B-Thinking (id `qwen3-vl-30b-a3b-thinking`, 31.1B params, MoE): 22.7 / 39 / 72.5
- LFM2.5-1.2B-Thinking (id `lfm2.5-1.2b-thinking`, 1.2B params): 1.8 / 2.4 / 3.7
- LFM3-VL-2B (id `lfm3-vl-2b`, 1.9B params): 2.3 / 3.3 / 5.4
- salamandra-7b-instruct (id `salamandra-7b-instruct`, 7.8B params): 6.4 / 10.5 / 18.9
- North-Micro-Vision-Instruct (id `north-micro-vision-instruct`, 2.5B params): 2.7 / 4 / 6.7
- Mage-VL (id `mage-vl`, 4.7B params): 4.3 / 6.8 / 11.9
- granite-4.0-3b-vision (id `granite-4.0-3b-vision`, 4.0B params): 3.8 / 5.9 / 10.2
- OLMo-2-0425-1B-Instruct (id `olmo-2-0425-1b-instruct`, 1.5B params): 2 / 2.8 / 4.4
- Hy-MT2-7B (id `hy-mt2-7b`, 8.0B params): 6.6 / 10.8 / 19.5
- LFM2-VL-450M (id `lfm2-vl-450m`, 0.5B params): 1.3 / 1.6 / 2
- granite-3.0-8b-instruct (id `granite-3.0-8b-instruct`, 8.2B params): 6.7 / 11 / 19.8
- Qwen2.5-32B (id `qwen2.5-32b`, 32.8B params): 23.8 / 41 / 76.4
- ERNIE-4.5-21B-A3B-PT (id `ernie-4.5-21b-a3b-pt`, 21.9B params, MoE): 16.3 / 27.8 / 51.5
- EXAONE-4.0-1.2B (id `exaone-4.0-1.2b`, 1.3B params): 1.9 / 2.6 / 3.9
- MiniCPM-SALA (id `minicpm-sala`, 9.5B params): 7.6 / 12.6 / 22.8
- Apertus-70B-Instruct-2509 (id `apertus-70b-instruct-2509`, 70.6B params): 50.2 / 87.3 / 163.4
- internlm2_5-7b-chat (id `internlm2-5-7b-chat`, 7.7B params): 6.4 / 10.5 / 18.8
- Moonlight-16B-A3B (id `moonlight-16b-a3b`, 16.0B params, MoE): 12.1 / 20.5 / 37.7
- Devstral-2-123B-Instruct-2512 (id `devstral-2-123b-instruct-2512`, 125.0B params): 88.2 / 153.8 / 288.6
- OLMo-2-1124-7B-Instruct (id `olmo-2-1124-7b-instruct`, 7.3B params): 6.1 / 9.9 / 17.8
- GLM-4-9B-0414 (id `glm-4-9b-0414`, 9.4B params): 7.6 / 12.5 / 22.6
- Yi-Coder-9B-Chat (id `yi-coder-9b-chat`, 8.8B params): 7.2 / 11.8 / 21.3
- Olmo-3-32B-Think (id `olmo-3-32b-think`, 32.2B params): 23.5 / 40.4 / 75.1
- Apertus-8B-2509 (id `apertus-8b-2509`, 8.1B params): 6.6 / 10.8 / 19.5
- MiniCPM4-8B (id `minicpm4-8b`, 8.2B params): 6.7 / 11 / 19.8
- LFM2-VL-1.6B (id `lfm2-vl-1.6b`, 1.6B params): 2.1 / 2.9 / 4.6
- L3-8B-Stheno-v3.2 (id `l3-8b-stheno-v3.2`, 8.0B params): 6.6 / 10.8 / 19.5
- MiniCPM4.1-8B (id `minicpm4.1-8b`, 8.2B params): 6.7 / 11 / 19.8
- HyperCLOVAX-SEED-Think-14B (id `hyperclovax-seed-think-14b`, 14.7B params): 11.3 / 19 / 34.9
- granite-4.1-30b (id `granite-4.1-30b`, 28.9B params): 21.1 / 36.3 / 67.4
- solar-pro-preview-instruct (id `solar-pro-preview-instruct`, 22.1B params): 16.4 / 28.1 / 51.9
- MiMo-7B-Base (id `mimo-7b-base`, 7.8B params): 6.5 / 10.6 / 19
- granite-4.0-1b (id `granite-4.0-1b`, 1.6B params): 2.1 / 3 / 4.8
- Qwen2.5-Coder-3B (id `qwen2.5-coder-3b`, 3.1B params): 3.2 / 4.8 / 8.1
- Hy-MT2-30B-A3B (id `hy-mt2-30b-a3b`, 30.1B params, MoE): 22 / 37.7 / 70.1
- Olmo-3-1125-32B (id `olmo-3-1125-32b`, 32.2B params): 23.5 / 40.4 / 75.1
- Fara-7B (id `fara-7b`, 8.3B params): 6.8 / 11.1 / 20.1
- Olmo-3.1-32B-Instruct (id `olmo-3.1-32b-instruct`, 32.2B params): 23.5 / 40.4 / 75.1
- Phi-4-reasoning-plus (id `phi-4-reasoning-plus`, 14.7B params): 11.2 / 18.9 / 34.7
- Nemotron-Labs-Diffusion-3B (id `nemotron-labs-diffusion-3b`, 3.8B params): 3.7 / 5.7 / 9.8
- K-EXAONE-2.0-750B-A37B (id `k-exaone-2.0-750b-a37b`, 749.4B params, MoE): 523.4 / 916.6 / 1724.5
- Youtu-LLM-2B (id `youtu-llm-2b`, 2.0B params): 2.4 / 3.4 / 5.5
- Llama-3.1-Nemotron-Nano-8B-v1 (id `llama-3.1-nemotron-nano-8b-v1`, 8.0B params): 6.6 / 10.8 / 19.5
- d1-2B (id `d1-2b`, 1.9B params): 2.3 / 3.3 / 5.4
- ERNIE-4.5-VL-28B-A3B-Thinking (id `ernie-4.5-vl-28b-a3b-thinking`, 29.7B params, MoE): 21.7 / 37.2 / 69.2
- MiniCPM-V-4.6-Thinking (id `minicpm-v-4.6-thinking`, 1.3B params): 1.9 / 2.6 / 4
- OLMo-2-0325-32B-Instruct (id `olmo-2-0325-32b-instruct`, 32.2B params): 23.5 / 40.4 / 75.1
- ERNIE-4.5-VL-28B-A3B-PT (id `ernie-4.5-vl-28b-a3b-pt`, 29.4B params, MoE): 21.5 / 36.9 / 68.6
- Mellum2-12B-A2.5B-Thinking (id `mellum2-12b-a2.5b-thinking`, 12.1B params, MoE): 9.5 / 15.8 / 28.9
- granite-4.0-350m (id `granite-4.0-350m`, 0.4B params): 1.2 / 1.4 / 1.8
- AliceAI-Foundation-80B-A3B-Base (id `aliceai-foundation-80b-a3b-base`, 81.3B params, MoE): 57.7 / 100.3 / 188
- YuE-s1-7B-anneal-en-cot (id `yue-s1-7b-anneal-en-cot`, 6.2B params): 5.3 / 8.6 / 15.3
- LFM2-24B-A2B (id `lfm2-24b-a2b`, 23.8B params, MoE): 17.6 / 30.1 / 55.8
- granite-3.1-2b-instruct (id `granite-3.1-2b-instruct`, 2.5B params): 2.8 / 4.1 / 6.8
- MiniMax-M1-80k (id `minimax-m1-80k`, 456.1B params, MoE): 319 / 558.3 / 1050
- LFM2-2.6B (id `lfm2-2.6b`, 2.6B params): 2.8 / 4.1 / 6.9
- granite-4.0-h-micro (id `granite-4.0-h-micro`, 3.2B params): 3.2 / 4.9 / 8.3
- WeVisOne-VL-4B (id `wevisone-vl-4b`, 4.8B params): 4.4 / 6.9 / 12.1
- Mellum-4b-base (id `mellum-4b-base`, 4.0B params): 3.8 / 5.9 / 10.2
- Phi-4-reasoning-vision-15B (id `phi-4-reasoning-vision-15b`, 15.1B params): 11.5 / 19.5 / 35.8
- ERNIE-4.5-0.3B-PT (id `ernie-4.5-0.3b-pt`, 0.4B params): 1.3 / 1.4 / 1.8
- Mellum2-12B-A2.5B-Instruct (id `mellum2-12b-a2.5b-instruct`, 12.1B params, MoE): 9.5 / 15.8 / 28.9
- Molmo2-O-7B (id `molmo2-o-7b`, 7.8B params): 6.4 / 10.5 / 18.8
- Llama-3.1-Tulu-3-8B (id `llama-3.1-tulu-3-8b`, 8.0B params): 6.6 / 10.8 / 19.5
- Llama-3_1-Nemotron-Ultra-253B-v1 (id `llama-3-1-nemotron-ultra-253b-v1`, 253.4B params): 177.7 / 310.6 / 583.8
- Hermes-3-Llama-3.2-3B (id `hermes-3-llama-3.2-3b`, 3.2B params): 3.2 / 4.9 / 8.4
- ReaderLM-v2 (id `readerlm-v2`, 1.5B params): 2.1 / 2.9 / 4.6
- Falcon3-7B-Instruct (id `falcon3-7b-instruct`, 7.5B params): 6.2 / 10.1 / 18.1
- Falcon3-10B-Instruct (id `falcon3-10b-instruct`, 10.3B params): 8.2 / 13.6 / 24.7
- Nemotron-Labs-Diffusion-8B-Base (id `nemotron-labs-diffusion-8b-base`, 8.5B params): 6.9 / 11.4 / 20.5
- MiMo-V2.6-Flash-MOPD (id `mimo-v2.6-flash-mopd`, 310.8B params, MoE): 217.7 / 380.7 / 715.7
- starvector-8b-im2svg (id `starvector-8b-im2svg`, 7.5B params): 6.2 / 10.2 / 18.3
- kanana-1.5-v-3b-instruct (id `kanana-1.5-v-3b-instruct`, 3.7B params): 3.6 / 5.5 / 9.4
- gemma-2-27b (id `gemma-2-27b`, 27.2B params): 20 / 34.3 / 63.6
- Llama-3.2-1B (id `llama-3.2-1b-nousresearch`, 1.2B params): 1.9 / 2.5 / 3.8
- WeVisOne-VL-2B (id `wevisone-vl-2b`, 2.4B params): 2.7 / 4 / 6.6
- LongCat-2.0 (id `longcat-2.0`, 1775.6B params): 1238.9 / 2170.5 / 4084.8
- AFM-4.5B (id `afm-4.5b`, 4.6B params): 4.2 / 6.6 / 11.6
- Olmo-3.1-32B-Think (id `olmo-3.1-32b-think`, 32.2B params): 23.5 / 40.4 / 75.1
- Olmo-Hybrid-7B (id `olmo-hybrid-7b`, 7.4B params): 6.2 / 10.1 / 18.1
- rnj-1-instruct (id `rnj-1-instruct`, 8.3B params): 6.8 / 11.2 / 20.1
- Meta-Llama-3.1-8B (id `meta-llama-3.1-8b`, 8.0B params): 6.6 / 10.8 / 19.5
- FastVLM-7B (id `fastvlm-7b`, 7.8B params): 6.4 / 10.5 / 18.9
- SmolLM-360M (id `smollm-360m`, 0.4B params): 1.3 / 1.4 / 1.8
- EuroLLM-22B-Instruct-2512 (id `eurollm-22b-instruct-2512`, 22.6B params): 16.8 / 28.7 / 53.1
- Phi-4-mini-flash-reasoning (id `phi-4-mini-flash-reasoning`, 3.9B params): 3.7 / 5.7 / 9.9
- MiMo-V2.6-Pro-MOPD (id `mimo-v2.6-pro-mopd`, 1024.2B params, MoE): 715.1 / 1252.5 / 2356.7
- Zamba2-7B-Instruct (id `zamba2-7b-instruct`, 7.4B params): 6.1 / 10 / 17.9
- codegeex4-all-9b (id `codegeex4-all-9b`, 9.4B params): 7.6 / 12.5 / 22.6
- OLMoE-1B-7B-0924-Instruct (id `olmoe-1b-7b-0924-instruct`, 6.9B params, MoE): 5.8 / 9.5 / 16.9
- MiniCPM4-0.5B (id `minicpm4-0.5b`, 0.4B params): 1.3 / 1.5 / 2
- Llama-4-Scout-17B-16E (id `llama-4-scout-17b-16e`, 108.6B params, MoE): 76.7 / 133.7 / 250.9
- c4ai-command-a-03-2025 (id `c4ai-command-a-03-2025`, 111.1B params): 78.4 / 136.7 / 256.4
- kanana-1.5-8b-instruct-2505 (id `kanana-1.5-8b-instruct-2505`, 8.0B params): 6.6 / 10.8 / 19.5
- Llama-3.1-Minitron-4B-Width-Base (id `llama-3.1-minitron-4b-width-base`, 4.5B params): 4.1 / 6.5 / 11.4
- Gemma-SEA-LION-v3-9B-IT (id `gemma-sea-lion-v3-9b-it`, 9.2B params): 7.4 / 12.3 / 22.3
- HY-Embodied-0.5 (id `hy-embodied-0.5`, 3.8B params): 3.6 / 5.6 / 9.7
- Marco-o1 (id `marco-o1`, 7.6B params): 6.3 / 10.3 / 18.5
- Falcon-H1R-7B (id `falcon-h1r-7b`, 7.6B params): 6.3 / 10.3 / 18.4
- LFM2-2.6B-Exp (id `lfm2-2.6b-exp`, 2.6B params): 2.8 / 4.1 / 6.9
- Mistral-NeMo-Minitron-8B-Base (id `mistral-nemo-minitron-8b-base`, 8.4B params): 6.9 / 11.3 / 20.4
- LFM2-700M (id `lfm2-700m`, 0.7B params): 1.5 / 1.9 / 2.7
- Intern-S1-mini (id `intern-s1-mini`, 8.5B params): 7 / 11.4 / 20.6
- UI-TARS-7B-DPO (id `ui-tars-7b-dpo`, 8.3B params): 6.8 / 11.1 / 20.1
- granite-4.0-h-1b (id `granite-4.0-h-1b`, 1.5B params): 2 / 2.8 / 4.4
- Seed-Coder-8B-Instruct (id `seed-coder-8b-instruct`, 8.3B params): 6.8 / 11.1 / 20
- LongCat-Flash-Lite (id `longcat-flash-lite`, 69.1B params): 49.2 / 85.4 / 159.9
- reka-flash-3 (id `reka-flash-3`, 20.9B params): 15.6 / 26.5 / 49.1
- Nanbeige4-3B-Thinking-2511 (id `nanbeige4-3b-thinking-2511`, 3.9B params): 3.7 / 5.8 / 10
- LFM2.5-1.2B-JP-202606 (id `lfm2.5-1.2b-jp-202606`, 1.2B params): 1.8 / 2.4 / 3.7
- Yi-Coder-9B (id `yi-coder-9b`, 8.8B params): 7.2 / 11.8 / 21.3
- Apriel-1.5-15b-Thinker (id `apriel-1.5-15b-thinker`, 14.9B params): 11.4 / 19.2 / 35.2
- Qwen2.5-Coder-14B (id `qwen2.5-coder-14b`, 14.8B params): 11.3 / 19 / 35
- Qwen2.5-Coder-32B (id `qwen2.5-coder-32b`, 32.8B params): 23.8 / 41 / 76.4
- MiMo-VL-7B-RL-2508 (id `mimo-vl-7b-rl-2508`, 8.3B params): 6.8 / 11.1 / 20.1
- granite-4.0-h-350m (id `granite-4.0-h-350m`, 0.3B params): 1.2 / 1.4 / 1.8
- HyperCLOVAX-SEED-Text-Instruct-0.5B (id `hyperclovax-seed-text-instruct-0.5b`, 0.6B params): 1.4 / 1.7 / 2.3
- IQuest-Coder-V1-40B-Loop-Instruct (id `iquest-coder-v1-40b-loop-instruct`, 39.8B params): 28.7 / 49.6 / 92.5
- circuit-sparsity (id `circuit-sparsity`, 0.4B params): 1.3 / 1.5 / 2
- Falcon3-1B-Instruct (id `falcon3-1b-instruct`, 1.7B params): 2.2 / 3 / 4.8
- AgentCPM-Report (id `agentcpm-report`, 8.2B params): 6.7 / 11 / 19.8
- Apriel-1.6-15b-Thinker (id `apriel-1.6-15b-thinker`, 14.9B params): 11.4 / 19.2 / 35.2
- OLMo-2-0325-32B (id `olmo-2-0325-32b`, 32.2B params): 23.5 / 40.4 / 75.1
- IQuest-Coder-V1-40B-Instruct (id `iquest-coder-v1-40b-instruct`, 39.8B params): 28.7 / 49.6 / 92.5
- granite-vision-3.2-2b (id `granite-vision-3.2-2b`, 3.0B params): 3.1 / 4.6 / 7.8
- aya-vision-8b (id `aya-vision-8b`, 8.6B params): 7 / 11.5 / 20.9
- Mage-VL-Cowork (id `mage-vl-cowork`, 4.7B params): 4.3 / 6.8 / 11.9
- Apriel-Nemotron-15b-Thinker (id `apriel-nemotron-15b-thinker`, 15.0B params): 11.4 / 19.3 / 35.4
- reader-lm-1.5b (id `reader-lm-1.5b`, 1.5B params): 2.1 / 2.9 / 4.6
- Zamba2-2.7B (id `zamba2-2.7b`, 2.7B params): 2.9 / 4.3 / 7.1
- HyperCLOVAX-SEED-Omni-8B (id `hyperclovax-seed-omni-8b`, 10.7B params): 8.5 / 14.1 / 25.7
- Molmo2-ER (id `molmo2-er`, 4.9B params): 4.4 / 6.9 / 12.2
- granite-3.2-2b-instruct (id `granite-3.2-2b-instruct`, 2.5B params): 2.8 / 4.1 / 6.8
- Zamba2-1.2B-instruct (id `zamba2-1.2b-instruct`, 1.2B params): 1.8 / 2.5 / 3.8
- MiniCPM5-1B-SFT (id `minicpm5-1b-sft`, 1.1B params): 1.8 / 2.3 / 3.5
- EXAONE-Deep-32B (id `exaone-deep-32b`, 32.0B params): 23.3 / 40.1 / 74.6
- Motif-3 (id `motif-3`, 314.8B params): 220.5 / 385.7 / 725.1
- MiniCPM4-MCP (id `minicpm4-mcp`, 8.2B params): 6.7 / 11 / 19.8
- Llama-3_1-Nemotron-51B-Instruct (id `llama-3-1-nemotron-51b-instruct`, 51.5B params): 36.9 / 63.9 / 119.5
- ERNIE-4.5-300B-A47B-PT (id `ernie-4.5-300b-a47b-pt`, 300.5B params, MoE): 210.5 / 368.1 / 692.1
- granite-switch-4.1-8b (id `granite-switch-4.1-8b`, 8.8B params): 7.1 / 11.7 / 21.2
- Apertus-70B-2509 (id `apertus-70b-2509`, 70.6B params): 50.2 / 87.3 / 163.4
- Olmo-3-7B-Instruct-SFT (id `olmo-3-7b-instruct-sft`, 7.3B params): 6.1 / 9.9 / 17.8
- MiMo-VL-7B-RL (id `mimo-vl-7b-rl`, 8.3B params): 6.8 / 11.1 / 20.1
- Youtu-VL-4B-Instruct (id `youtu-vl-4b-instruct`, 5.3B params): 4.7 / 7.5 / 13.3
- Seed-Coder-8B-Reasoning (id `seed-coder-8b-reasoning`, 8.3B params): 6.8 / 11.1 / 20
- UI-TARS-72B-DPO (id `ui-tars-72b-dpo`, 73.4B params): 52.2 / 90.7 / 169.8
- Spark-X2.6-1.7B (id `spark-x2.6-1.7b`, 1.7B params): 2.2 / 3.1 / 4.9
- Falcon3-3B-Instruct (id `falcon3-3b-instruct`, 3.2B params): 3.3 / 4.9 / 8.4
- Falcon-H1-1.5B-Deep-Instruct (id `falcon-h1-1.5b-deep-instruct`, 1.6B params): 2.1 / 2.9 / 4.6
- kanana-2-3b-instruct (id `kanana-2-3b-instruct`, 3.5B params): 3.4 / 5.3 / 9.1
- GLM-Z1-Rumination-32B-0414 (id `glm-z1-rumination-32b-0414`, 33.1B params): 24.1 / 41.5 / 77.2
- EuroLLM-9B-Instruct-2512 (id `eurollm-9b-instruct-2512`, 9.2B params): 7.4 / 12.2 / 22.1
- WeDLM-8B-Instruct (id `wedlm-8b-instruct`, 8.2B params): 6.7 / 11 / 19.8
- Falcon-H1-7B-Instruct (id `falcon-h1-7b-instruct`, 7.6B params): 6.3 / 10.3 / 18.4
- granite-vision-3.1-2b-preview (id `granite-vision-3.1-2b-preview`, 3.0B params): 3.1 / 4.6 / 7.8
- GLM-Z1-9B-0414 (id `glm-z1-9b-0414`, 9.4B params): 7.6 / 12.5 / 22.6
- UI-Venus-1.5-2B (id `ui-venus-1.5-2b`, 2.4B params): 2.7 / 4 / 6.6
- Instella-MoE-16B-A3B-Think (id `instella-moe-16b-a3b-think`, 15.9B params, MoE): 12.1 / 20.4 / 37.5
- granite-switch-4.1-30b (id `granite-switch-4.1-30b`, 30.0B params): 21.9 / 37.6 / 70
- MiniCPM5-2B-SFT (id `minicpm5-2b-sft`, 2.5B params): 2.8 / 4.1 / 6.8
- Apriel-5B-Instruct (id `apriel-5b-instruct`, 4.8B params): 4.4 / 6.9 / 12.1
- kanana-1.5-2.1b-instruct-2505 (id `kanana-1.5-2.1b-instruct-2505`, 2.3B params): 2.6 / 3.8 / 6.3
- LongCat-Flash-Thinking (id `longcat-flash-thinking`, 561.9B params): 392.7 / 687.5 / 1293.3
- FastVLM-1.5B (id `fastvlm-1.5b`, 1.9B params): 2.3 / 3.3 / 5.4
- Falcon-H1-34B-Instruct (id `falcon-h1-34b-instruct`, 33.6B params): 24.5 / 42.1 / 78.4
- granite-3.0-2b-instruct (id `granite-3.0-2b-instruct`, 2.6B params): 2.8 / 4.2 / 7.1
- MediPhi-Instruct (id `mediphi-instruct`, 3.8B params): 3.7 / 5.7 / 9.8
- LFM2-2.6B-Transcript (id `lfm2-2.6b-transcript`, 2.6B params): 2.8 / 4.1 / 6.9
- Falcon-H1-3B-Instruct (id `falcon-h1-3b-instruct`, 3.1B params): 3.2 / 4.8 / 8.2
- bitnet-b1.58-2B-4T-bf16 (id `bitnet-b1.58-2b-4t-bf16`, 2.4B params): 2.7 / 3.9 / 6.5
- Falcon-H1-0.5B-Instruct (id `falcon-h1-0.5b-instruct`, 0.5B params): 1.4 / 1.6 / 2.2
- LongCat-Flash-Thinking-2601 (id `longcat-flash-thinking-2601`, 561.9B params): 392.7 / 687.5 / 1293.3
- EXAONE-Deep-7.8B (id `exaone-deep-7.8b`, 7.8B params): 6.5 / 10.6 / 19
- glm-4-9b-chat-hf (id `glm-4-9b-chat-hf`, 9.4B params): 7.6 / 12.5 / 22.6
- GLM-4.1V-9B-Base (id `glm-4.1v-9b-base`, 10.3B params): 8.2 / 13.6 / 24.7
- UI-TARS-7B-SFT (id `ui-tars-7b-sft`, 8.3B params): 6.8 / 11.1 / 20.1
- sarvam-1-v0.5 (id `sarvam-1-v0.5`, 2.5B params): 2.7 / 4.1 / 6.8
- Llama-3.1-Tulu-3-8B-DPO (id `llama-3.1-tulu-3-8b-dpo`, 8.0B params): 6.6 / 10.8 / 19.5
- MiMo-Embodied-7B (id `mimo-embodied-7b`, 8.3B params): 6.8 / 11.1 / 20.1
- OLMo-1B-0724-hf (id `olmo-1b-0724-hf`, 1.3B params): 1.9 / 2.6 / 3.9
- internlm2_5-20b-chat (id `internlm2-5-20b-chat`, 19.9B params): 14.8 / 25.3 / 46.7
- Glyph (id `glyph`, 10.3B params): 8.2 / 13.6 / 24.7
- Mellum2-12B-A2.5B-Thinking-SFT (id `mellum2-12b-a2.5b-thinking-sft`, 12.1B params, MoE): 9.5 / 15.8 / 28.9
- Zamba2-2.7B-instruct (id `zamba2-2.7b-instruct`, 2.7B params): 2.9 / 4.3 / 7.1
- Olmo-3-7B-Think-SFT (id `olmo-3-7b-think-sft`, 7.3B params): 6.1 / 9.9 / 17.8
- Hemmingway-1.5 (id `hemmingway-1.5`, 26.9B params): 19.8 / 33.9 / 62.9
- POINTS-Reader (id `points-reader`, 4.1B params): 3.9 / 6 / 10.5
- internlm2_5-1_8b-chat (id `internlm2-5-1-8b-chat`, 1.9B params): 2.3 / 3.3 / 5.3
- ERNIE-4.5-VL-424B-A47B-PT (id `ernie-4.5-vl-424b-a47b-pt`, 423.5B params, MoE): 296.3 / 518.5 / 975.1
- MiniCPM5-2B-Midtrain (id `minicpm5-2b-midtrain`, 2.5B params): 2.8 / 4.1 / 6.8
- Hermes-2-Theta-Llama-3-70B (id `hermes-2-theta-llama-3-70b`, 70.6B params): 50.2 / 87.2 / 163.3
- EXAONE-Deep-2.4B (id `exaone-deep-2.4b`, 2.4B params): 2.7 / 3.9 / 6.5
- Stable-DiffCoder-8B-Instruct (id `stable-diffcoder-8b-instruct`, 8.3B params): 6.8 / 11.1 / 20
- UI-Venus-1.5-8B (id `ui-venus-1.5-8b`, 8.8B params): 7.1 / 11.7 / 21.2
- SeaLLMs-v3-7B-Chat (id `seallms-v3-7b-chat`, 7.6B params): 6.3 / 10.3 / 18.5
- SmolVLM-Base (id `smolvlm-base`, 2.2B params): 2.6 / 3.7 / 6.2
- UI-TARS-2B-SFT (id `ui-tars-2b-sft`, 2.4B params): 2.7 / 4 / 6.6
- EXAONE-4.0.1-32B (id `exaone-4.0.1-32b`, 32.0B params): 23.3 / 40.1 / 74.6
- Bolmo-1B (id `bolmo-1b`, 1.5B params): 2 / 2.8 / 4.4
- ALIA-40b (id `alia-40b`, 40.4B params): 29.2 / 50.4 / 94
- INTELLECT-1-Instruct (id `intellect-1-instruct`, 10.2B params): 8.1 / 13.5 / 24.5
- salamandra-2b (id `salamandra-2b`, 2.3B params): 2.6 / 3.8 / 6.2
- kanana-nano-2.1b-instruct (id `kanana-nano-2.1b-instruct`, 2.1B params): 2.5 / 3.6 / 5.8
- Olmo-3.1-32B-Instruct-DPO (id `olmo-3.1-32b-instruct-dpo`, 32.2B params): 23.5 / 40.4 / 75.1
- salamandra-7b (id `salamandra-7b`, 7.8B params): 6.4 / 10.5 / 18.9
- Llama-3.1-Tulu-3.1-8B (id `llama-3.1-tulu-3.1-8b`, 8.0B params): 6.6 / 10.8 / 19.5
- salamandra-2b-instruct (id `salamandra-2b-instruct`, 2.3B params): 2.6 / 3.8 / 6.2
- Olmo-3-7B-Instruct-DPO (id `olmo-3-7b-instruct-dpo`, 7.3B params): 6.1 / 9.9 / 17.8
- MiMo-7B-RL-0530 (id `mimo-7b-rl-0530`, 7.8B params): 6.5 / 10.6 / 19
- Youtu-LLM-A7B-Base (id `youtu-llm-a7b-base`, 118.4B params, MoE): 83.5 / 145.7 / 273.3
- OLMo-7B-0724-hf (id `olmo-7b-0724-hf`, 6.9B params): 5.8 / 9.4 / 16.8
- EuroLLM-22B-2512 (id `eurollm-22b-2512`, 22.6B params): 16.8 / 28.7 / 53.1
- Arcee-Spark (id `arcee-spark`, 7.6B params): 6.3 / 10.3 / 18.5
- CapRL-3B (id `caprl-3b`, 3.8B params): 3.6 / 5.6 / 9.6
- PaCoRe-8B (id `pacore-8b`, 8.2B params): 6.7 / 11 / 19.8
- Falcon-H1-1.5B-Instruct (id `falcon-h1-1.5b-instruct`, 1.6B params): 2.1 / 2.9 / 4.6
- UI-Venus-1.5-30B-A3B (id `ui-venus-1.5-30b-a3b`, 31.1B params, MoE): 22.7 / 39 / 72.5
- internlm2_5-7b-chat-1m (id `internlm2-5-7b-chat-1m`, 7.7B params): 6.4 / 10.5 / 18.8
- MiMo-VL-7B-SFT (id `mimo-vl-7b-sft`, 8.3B params): 6.8 / 11.1 / 20.1
- Olmo-3-7B-RL-Zero-Math (id `olmo-3-7b-rl-zero-math`, 7.3B params): 6.1 / 9.9 / 17.8
- MiMo-7B-SFT (id `mimo-7b-sft`, 7.8B params): 6.5 / 10.6 / 19
- glm-edge-1.5b-chat (id `glm-edge-1.5b-chat`, 1.6B params): 2.1 / 2.9 / 4.7
- MediPhi (id `mediphi`, 3.8B params): 3.7 / 5.7 / 9.8
- Olmo-3.1-7B-RL-Zero-Math (id `olmo-3.1-7b-rl-zero-math`, 7.3B params): 6.1 / 9.9 / 17.8
- Olmo-3-7B-Think-DPO (id `olmo-3-7b-think-dpo`, 7.3B params): 6.1 / 9.9 / 17.8
- JanusCoder-14B (id `januscoder-14b`, 14.8B params): 11.3 / 19 / 35
- Yi-Coder-1.5B-Chat (id `yi-coder-1.5b-chat`, 1.5B params): 2 / 2.8 / 4.4
- INTELLECT-1 (id `intellect-1`, 10.2B params): 8.1 / 13.5 / 24.5
- Motif-2.6B (id `motif-2.6b`, 2.6B params): 2.8 / 4.2 / 7
- MediPhi-Clinical (id `mediphi-clinical`, 3.8B params): 3.7 / 5.7 / 9.8
- kanana-2-1.3b-instruct (id `kanana-2-1.3b-instruct`, 1.3B params): 1.9 / 2.6 / 4
- LongWriter-llama3.1-8b (id `longwriter-llama3.1-8b`, 8.0B params): 6.6 / 10.8 / 19.5
- salamandra-7b-instruct-2606 (id `salamandra-7b-instruct-2606`, 7.8B params): 6.4 / 10.5 / 18.9
- granite-vision-3.3-2b-chart2csv-preview (id `granite-vision-3.3-2b-chart2csv-preview`, 3.0B params): 3.1 / 4.6 / 7.8
- Seed-OSS-36B-Base-woSyn (id `seed-oss-36b-base-wosyn`, 36.2B params): 26.2 / 45.2 / 84.1
- Qwen3.8-27B-Longbook27-v2 (id `qwen3.8-27b-longbook27-v2`, 27.1B params): 19.9 / 34.1 / 63.3
- Falcon-H1-Tiny-R-0.6B (id `falcon-h1-tiny-r-0.6b`, 0.6B params): 1.4 / 1.8 / 2.4
- granite-swash-2b (id `granite-swash-2b`, 2.1B params): 2.5 / 3.6 / 5.9
- Falcon-E-3B-Instruct (id `falcon-e-3b-instruct`, 0.9B params): 1.6 / 2.1 / 3
- MediPhi-PubMed (id `mediphi-pubmed`, 3.8B params): 3.7 / 5.7 / 9.8
- glm-edge-4b-chat (id `glm-edge-4b-chat`, 4.3B params): 4 / 6.3 / 11
- ZAYA1-base (id `zaya1-base`, 8.8B params): 7.2 / 11.8 / 21.3
- AMD-OLMo-1B (id `amd-olmo-1b`, 1.2B params): 1.8 / 2.4 / 3.7
- LLaDA-MoE-7B-A1B-Base (id `llada-moe-7b-a1b-base`, 7.4B params, MoE): 6.1 / 10 / 17.9
- Meta-Llama-3.1-70B-Instruct (id `meta-llama-3.1-70b-instruct`, 70.6B params): 50.2 / 87.2 / 163.3
- CapRL-Qwen3VL-4B (id `caprl-qwen3vl-4b`, 4.4B params): 4.1 / 6.4 / 11.2
- UI-Venus-Ground-7B (id `ui-venus-ground-7b`, 8.3B params): 6.8 / 11.1 / 20.1
- DeepHermes-3-Llama-3-3B-Preview (id `deephermes-3-llama-3-3b-preview`, 3.2B params): 3.2 / 4.9 / 8.4
- EuroMoE-2.6B-A0.6B-Instruct-Preview (id `euromoe-2.6b-a0.6b-instruct-preview`, 2.6B params, MoE): 2.8 / 4.2 / 7
- Apertus-v1.1-0.5B (id `apertus-v1.1-0.5b`, 0.4B params): 1.3 / 1.5 / 2
- Mellum2.1-12B-A2.5B-Thinking-MTP (id `mellum2.1-12b-a2.5b-thinking-mtp`, 12.1B params, MoE): 9.5 / 15.8 / 28.9
- DeepSeek-V4-Flash-0731 (id `deepseek-v4-flash-0731`, 304.2B params, MoE): 213.1 / 372.7 / 700.6
- Inkling (id `inkling`, 952.4B params, MoE): 665 / 1164.7 / 2191.5
- MiniMax-M2.5 (id `minimax-m2.5`, 228.7B params, MoE): 160.4 / 280.4 / 527
- GLM-5.2 (id `glm-5.2`, 753.3B params, MoE): 526.2 / 921.5 / 1733.7
- GLM-5.1 (id `glm-5.1`, 753.9B params, MoE): 526.6 / 922.1 / 1734.9
- Hy3 (id `hy3`, 298.8B params, MoE): 209.3 / 366.1 / 688.2
- Qwen3-Coder-Next (id `qwen3-coder-next`, 79.7B params, MoE): 56.5 / 98.4 / 184.3
- Laguna-S-2.1 (id `laguna-s-2.1`, 117.6B params, MoE): 83 / 144.6 / 271.4
- Inkling-Small (id `inkling-small`, 266.0B params, MoE): 186.4 / 326 / 612.7
- MiniMax-M2.7 (id `minimax-m2.7`, 228.7B params, MoE): 160.4 / 280.4 / 527
- Laguna-XS.2 (id `laguna-xs.2`, 33.4B params, MoE): 24.3 / 41.9 / 77.9
- Step-3.5-Flash (id `step-3.5-flash`, 199.4B params, MoE): 140 / 244.6 / 459.6
- DeepSeek-V4-Flash-Vision-Exp (id `deepseek-v4-flash-vision-exp`, 304.6B params, MoE): 213.4 / 373.2 / 701.7
- GLM-4.7 (id `glm-4.7`, 358.3B params, MoE): 250.8 / 438.8 / 825.2
- DeepSeek-V4-Pro-0813 (id `deepseek-v4-pro-0813`, 1650.5B params, MoE): 1151.7 / 2017.7 / 3797.1
- DeepSeek-Coder-V2-Lite-Instruct (id `deepseek-coder-v2-lite-instruct`, 15.7B params, MoE): 12 / 20.2 / 37.1
- Hy3-preview (id `hy3-preview`, 298.8B params, MoE): 209.3 / 366.1 / 688.2
- MiniMax-M2.1 (id `minimax-m2.1`, 228.7B params, MoE): 160.4 / 280.4 / 527
- Solar-Open-100B (id `solar-open-100b`, 102.7B params, MoE): 72.6 / 126.4 / 237.1
- Mistral-Small-4-119B-2603 (id `mistral-small-4-119b-2603`, 119.4B params, MoE): 84.2 / 146.9 / 275.6
- DeepSeek-V3.1-Base (id `deepseek-v3.1-base`, 684.5B params, MoE): 478.2 / 837.4 / 1575.4
- Agents-A1 (id `agents-a1`, 35.1B params, MoE): 25.5 / 43.9 / 81.7
- Ling-3.0-tiny (id `ling-3.0-tiny`, 7.9B params, MoE): 6.5 / 10.6 / 19.2
- OLMoE-1B-7B-0924 (id `olmoe-1b-7b-0924`, 6.9B params, MoE): 5.8 / 9.5 / 16.9
- Hy4-preview (id `hy4-preview`, 780.0B params, MoE): 544.8 / 954 / 1794.9
- Step-3.7-Flash (id `step-3.7-flash`, 201.4B params, MoE): 141.4 / 247 / 464.1
- LLaDA2.1-mini (id `llada2.1-mini`, 16.3B params, MoE): 12.3 / 20.9 / 38.4
- Ling-3.0-flash (id `ling-3.0-flash`, 127.5B params, MoE): 89.9 / 156.8 / 294.2
- granite-3.1-3b-a800m-instruct (id `granite-3.1-3b-a800m-instruct`, 3.3B params, MoE): 3.3 / 5 / 8.6
- LLaDA2.0-mini (id `llada2.0-mini`, 16.3B params, MoE): 12.3 / 20.9 / 38.4
- command-a-plus-05-2026-bf16 (id `command-a-plus-05-2026-bf16`, 218.8B params, MoE): 153.5 / 268.3 / 504.1
- OLMoE-1B-7B-0125-Instruct (id `olmoe-1b-7b-0125-instruct`, 6.9B params, MoE): 5.8 / 9.5 / 16.9
- GLM-5.3-BF16 (id `glm-5.3-bf16`, 753.3B params, MoE): 526.2 / 921.5 / 1733.7
- Trinity-Mini (id `trinity-mini`, 26.1B params, MoE): 19.2 / 32.9 / 61.1
- Phi-tiny-MoE-instruct (id `phi-tiny-moe-instruct`, 3.8B params, MoE): 3.6 / 5.6 / 9.6
- Ling-3.0-flash-VL (id `ling-3.0-flash-vl`, 124.8B params, MoE): 88 / 153.5 / 288.2
- granite-3.1-1b-a400m-instruct (id `granite-3.1-1b-a400m-instruct`, 1.3B params, MoE): 1.9 / 2.6 / 4.1
- Solar-Open2-250B (id `solar-open2-250b`, 250.3B params, MoE): 175.5 / 306.8 / 576.7
- North-Mini-Code-1.0 (id `north-mini-code-1.0`, 30.5B params, MoE): 22.3 / 38.2 / 71.1
- AI21-Jamba-Reasoning-3B (id `ai21-jamba-reasoning-3b`, 3.2B params): 3.2 / 4.9 / 8.4
- KAT-Coder-V2.5-Dev (id `kat-coder-v2.5-dev`, 34.7B params, MoE): 25.2 / 43.4 / 80.7
- DeepSeek-R1-Zero (id `deepseek-r1-zero`, 684.5B params, MoE): 478.2 / 837.4 / 1575.3
- DeepSeek-V2.5 (id `deepseek-v2.5`, 235.7B params, MoE): 165.4 / 289 / 543.2
- Ling-2.6-flash (id `ling-2.6-flash`, 107.5B params, MoE): 75.9 / 132.3 / 248.2
- Nex-N2.5-mini (id `nex-n2.5-mini`, 35.1B params, MoE): 25.5 / 43.9 / 81.7
- DeepSeek-V3.2-Speciale (id `deepseek-v3.2-speciale`, 685.4B params, MoE): 478.9 / 838.5 / 1577.4
- DeepSeek-Coder-V2-Instruct (id `deepseek-coder-v2-instruct`, 235.7B params, MoE): 165.4 / 289 / 543.2
- Intern-S1 (id `intern-s1`, 240.7B params, MoE): 168.8 / 295.1 / 554.6
- GLM-4.6V (id `glm-4.6v`, 107.7B params, MoE): 76.1 / 132.6 / 248.7
- sarvam-105b (id `sarvam-105b`, 106.0B params, MoE): 74.9 / 130.6 / 244.9
- Xing4.0-29B-Intent (id `xing4.0-29b-intent`, 31.2B params, MoE): 22.8 / 39.1 / 72.8
- Ling-1T (id `ling-1t`, 999.7B params, MoE): 698 / 1222.5 / 2300.3
- Intern-S1-Pro (id `intern-s1-pro`, 917.0B params, MoE): 640.4 / 1121.5 / 2110.2
- granite-4.0-tiny-preview (id `granite-4.0-tiny-preview`, 6.7B params, MoE): 5.7 / 9.2 / 16.3
- Nex-N2.5-Pro (id `nex-n2.5-pro`, 396.8B params, MoE): 277.6 / 485.8 / 913.6
- Ling-3.1-flash (id `ling-3.1-flash`, 572.1B params, MoE): 399.9 / 700.1 / 1316.9
- sarvam-30b (id `sarvam-30b`, 32.2B params, MoE): 23.4 / 40.3 / 75
- Ling-mini-2.0 (id `ling-mini-2.0`, 16.3B params, MoE): 12.3 / 20.9 / 38.4
- Ling-flash-2.0 (id `ling-flash-2.0`, 102.9B params, MoE): 72.7 / 126.7 / 237.6
- Step-5-Preview (id `step-5-preview`, 604.3B params, MoE): 422.3 / 739.4 / 1391
- DeepSeek-V2.5-1210 (id `deepseek-v2.5-1210`, 235.7B params, MoE): 165.4 / 289 / 543.2
- maple-preview (id `maple-preview`, 20.2B params, MoE): 15.1 / 25.7 / 47.5
- Trinity-Large-Thinking (id `trinity-large-thinking`, 398.6B params, MoE): 278.9 / 488.1 / 917.9
- DeepSeek-V2-Chat-0628 (id `deepseek-v2-chat-0628`, 235.7B params, MoE): 165.4 / 289 / 543.2
- AI21-Jamba2-3B (id `ai21-jamba2-3b`, 3.0B params): 3.1 / 4.7 / 8
- Kimi-K3.1-Pro (id `kimi-k3.1-pro`, 2780.1B params, MoE): 1939.3 / 3397.9 / 6395.3
- Hymba-1.5B-Instruct (id `hymba-1.5b-instruct`, 1.5B params): 2.1 / 2.9 / 4.5
- Nex-N2-Pro (id `nex-n2-pro`, 396.8B params, MoE): 277.6 / 485.8 / 913.6
- Ring-2.5-1T (id `ring-2.5-1t`, 1012.2B params, MoE): 706.7 / 1237.8 / 2329.1
- OLMoE-1B-7B-0125 (id `olmoe-1b-7b-0125`, 6.9B params, MoE): 5.8 / 9.5 / 16.9
- Intern-S2-Preview (id `intern-s2-preview`, 36.1B params, MoE): 26.2 / 45.1 / 84
- Ring-1T (id `ring-1t`, 999.7B params, MoE): 698 / 1222.5 / 2300.3
- Trinity-Large-Preview (id `trinity-large-preview`, 398.6B params, MoE): 278.9 / 488.1 / 917.9
- Ling-3.0-flash-Fin (id `ling-3.0-flash-fin`, 127.5B params, MoE): 89.9 / 156.8 / 294.2
- Nex-N2-mini (id `nex-n2-mini`, 35.1B params, MoE): 25.5 / 43.9 / 81.7
- Ling-3.1-flash-1019 (id `ling-3.1-flash-1019`, 572.1B params, MoE): 399.9 / 700.1 / 1316.9
- Trinity-Nano-Preview (id `trinity-nano-preview`, 6.1B params, MoE): 5.3 / 8.5 / 15.1
- Ring-mini-2.0 (id `ring-mini-2.0`, 16.3B params, MoE): 12.3 / 20.9 / 38.4
- Phi-mini-MoE-instruct (id `phi-mini-moe-instruct`, 7.6B params, MoE): 6.3 / 10.3 / 18.6
- Ling-lite (id `ling-lite`, 16.8B params, MoE): 12.7 / 21.5 / 39.6
- LLaDA2.2-mini (id `llada2.2-mini`, 16.3B params, MoE): 12.3 / 20.9 / 38.4
- FlexOlmo-7x7B-1T (id `flexolmo-7x7b-1t`, 33.3B params): 24.2 / 41.7 / 77.5
- DeepSeek-Coder-V2-Instruct-0724 (id `deepseek-coder-v2-instruct-0724`, 235.7B params, MoE): 165.4 / 289 / 543.2
- kanana-2-30b-a3b-instruct-2601 (id `kanana-2-30b-a3b-instruct-2601`, 30.7B params, MoE): 22.4 / 38.5 / 71.5
- LLaDA2.1-flash (id `llada2.1-flash`, 102.9B params, MoE): 72.7 / 126.7 / 237.6
- Ring-flash-2.0 (id `ring-flash-2.0`, 102.9B params, MoE): 72.7 / 126.7 / 237.6
- granite-3.0-1b-a400m-instruct (id `granite-3.0-1b-a400m-instruct`, 1.3B params, MoE): 1.9 / 2.6 / 4.1
- GRIN-MoE (id `grin-moe`, 41.9B params, MoE): 30.2 / 52.2 / 97.3
- LLaDA2.0-mini-preview (id `llada2.0-mini-preview`, 16.3B params, MoE): 12.3 / 20.9 / 38.4
- kanana-2-30b-a3b-instruct (id `kanana-2-30b-a3b-instruct`, 30.7B params, MoE): 22.4 / 38.5 / 71.5
- Ling-lite-1.5-2507 (id `ling-lite-1.5-2507`, 16.8B params, MoE): 12.7 / 21.5 / 39.6
- LLaDA2.2-flash (id `llada2.2-flash`, 102.9B params, MoE): 72.7 / 126.7 / 237.6
- Trinity-Large-TrueBase (id `trinity-large-truebase`, 398.6B params, MoE): 278.9 / 488.1 / 917.9
- kanana-2-30b-a3b-thinking-2601 (id `kanana-2-30b-a3b-thinking-2601`, 30.7B params, MoE): 22.4 / 38.5 / 71.5
- Ling-lite-1.5 (id `ling-lite-1.5`, 16.8B params, MoE): 12.7 / 21.5 / 39.6
- Ring-mini-linear-2.0 (id `ring-mini-linear-2.0`, 16.4B params, MoE): 12.5 / 21.1 / 38.8
- Trinity-Large-Base (id `trinity-large-base`, 398.6B params, MoE): 278.9 / 488.1 / 917.9
- AI21-Jamba2-Mini (id `ai21-jamba2-mini`, 51.6B params, MoE): 37 / 64 / 119.6
- granite-swash-3b-a600m (id `granite-swash-3b-a600m`, 3.0B params, MoE): 3.1 / 4.7 / 7.9
- Ling-2.5-1T (id `ling-2.5-1t`, 1012.2B params, MoE): 706.7 / 1237.8 / 2329.1
- Intern-S2-397B (id `intern-s2-397b`, 403.4B params, MoE): 282.3 / 493.9 / 928.9
- Ling-Coder-lite (id `ling-coder-lite`, 16.8B params, MoE): 12.7 / 21.5 / 39.6
- Ling-plus (id `ling-plus`, 292.5B params, MoE): 205 / 358.4 / 673.8
- Step-3.5-Flash-Base-Midtrain (id `step-3.5-flash-base-midtrain`, 197.8B params, MoE): 138.9 / 242.6 / 455.9
- Ring-lite (id `ring-lite`, 16.8B params, MoE): 12.7 / 21.5 / 39.6
- Trinity-Nano-Base (id `trinity-nano-base`, 6.1B params, MoE): 5.3 / 8.5 / 15.1
- Ring-lite-distill-preview (id `ring-lite-distill-preview`, 16.8B params, MoE): 12.7 / 21.5 / 39.6
- kanana-1.5-15.7b-a3b-instruct (id `kanana-1.5-15.7b-a3b-instruct`, 15.7B params, MoE): 11.9 / 20.2 / 37.1
- Rubicon-Preview (id `rubicon-preview`, 30.5B params, MoE): 22.3 / 38.3 / 71.2
- Axion-1.0-35B (id `axion-1.0-35b`, 35.1B params, MoE): 25.5 / 43.9 / 81.7
- LFM3-Omni-2B (id `lfm3-omni-2b`, 2.2B params): 2.5 / 3.6 / 6
- Clover-35B-A3B-Preview (id `clover-35b-a3b-preview`, 34.7B params, MoE): 25.2 / 43.4 / 80.7
- SenseNova-Loop-9B-Preview (id `sensenova-loop-9b-preview`, 9.7B params): 7.7 / 12.8 / 23.2
- Arete-27B (id `arete-27b`, 27.4B params): 20.1 / 34.4 / 63.9
- ERNIE-5.5-Flash-Base (id `ernie-5.5-flash-base`, 286.2B params, MoE): 200.6 / 350.8 / 659.4
- Groot-19B-A3B-261023 (id `groot-19b-a3b-261023`, 19.4B params, MoE): 14.5 / 24.7 / 45.7
- Apex-35B-V1.0 (id `apex-35b-v1.0`, 36.0B params, MoE): 26.1 / 44.9 / 83.7
- S5-nano (id `s5-nano`, 4.7B params): 4.2 / 6.7 / 11.7
- NorthCUA-Pro (id `northcua-pro`, 396.8B params, MoE): 277.6 / 485.8 / 913.6
- NorthCUA-Mini (id `northcua-mini`, 35.1B params, MoE): 25.5 / 43.9 / 81.7
- NorthCUA-Nano (id `northcua-nano`, 9.4B params): 7.6 / 12.5 / 22.6
- LongCat-2.5 (id `longcat-2.5`, 1779.8B params, MoE): 1241.8 / 2175.6 / 4094.4

## Hardware (usable GB, bandwidth GB/s, list price)
- GeForce RTX 3050 8GB (id `rtx-3050-8gb`, gpu): 8 GB usable, 224 GB/s, $249 list
- GeForce RTX 3060 12GB (id `rtx-3060-12gb`, gpu): 12 GB usable, 360 GB/s, $329 list
- GeForce RTX 3060 Ti (id `rtx-3060-ti`, gpu): 8 GB usable, 448 GB/s, $399 list
- GeForce RTX 3070 (id `rtx-3070`, gpu): 8 GB usable, 448 GB/s, $499 list
- GeForce RTX 3080 10GB (id `rtx-3080-10gb`, gpu): 10 GB usable, 760 GB/s, $699 list
- GeForce RTX 3080 Ti (id `rtx-3080-ti`, gpu): 12 GB usable, 912 GB/s, $1199 list
- GeForce RTX 3090 (id `rtx-3090`, gpu): 24 GB usable, 936 GB/s, $1499 list
- GeForce RTX 3090 Ti (id `rtx-3090-ti`, gpu): 24 GB usable, 1008 GB/s, $1999 list
- GeForce RTX 4060 (id `rtx-4060`, gpu): 8 GB usable, 272 GB/s, $299 list
- GeForce RTX 4060 Ti 16GB (id `rtx-4060-ti-16gb`, gpu): 16 GB usable, 288 GB/s, $499 list
- GeForce RTX 4070 (id `rtx-4070`, gpu): 12 GB usable, 504 GB/s, $599 list
- GeForce RTX 4070 Super (id `rtx-4070-super`, gpu): 12 GB usable, 504 GB/s, $599 list
- GeForce RTX 4070 Ti Super (id `rtx-4070-ti-super`, gpu): 16 GB usable, 672 GB/s, $799 list
- GeForce RTX 4080 Super (id `rtx-4080-super`, gpu): 16 GB usable, 736 GB/s, $999 list
- GeForce RTX 4090 (id `rtx-4090`, gpu): 24 GB usable, 1008 GB/s, $1599 list
- GeForce RTX 5070 (id `rtx-5070`, gpu): 12 GB usable, 672 GB/s, $549 list
- GeForce RTX 5080 (id `rtx-5080`, gpu): 16 GB usable, 960 GB/s, $999 list
- GeForce RTX 5090 (id `rtx-5090`, gpu): 32 GB usable, 1792 GB/s, $1999 list
- Radeon RX 7600 XT (id `rx-7600-xt`, gpu): 16 GB usable, 288 GB/s, $329 list
- Radeon RX 7800 XT (id `rx-7800-xt`, gpu): 16 GB usable, 624 GB/s, $499 list
- Radeon RX 7900 XT (id `rx-7900-xt`, gpu): 20 GB usable, 800 GB/s, $899 list
- Radeon RX 7900 XTX (id `rx-7900-xtx`, gpu): 24 GB usable, 960 GB/s, $999 list
- Intel Arc B580 (id `arc-b580`, gpu): 12 GB usable, 456 GB/s, $249 list
- Intel Arc Pro B70 32GB (id `arc-pro-b70`, gpu): 32 GB usable, 608 GB/s, $949 list
- Mac mini M4 16GB (id `mac-mini-m4-16`, unified): 12 GB usable, 120 GB/s, $789 B&H sale, 16GB/256GB MU9D3LL/A (was $899 Apple list; discontinued at Apple, replaced by M6; was $799 on 2026-10-29)
- Mac mini M6 16GB (id `mac-mini-m6-16`, unified): 12 GB usable, 153 GB/s, $899 Apple list price, M6 12C CPU/12C GPU, 16GB/256GB (MHQK4LL/A)
- Mac mini M6 24GB (id `mac-mini-m6-24`, unified): 18 GB usable, 170 GB/s, $1299 Apple list price, M6 12C CPU/12C GPU, 24GB/512GB (MHQM4LL/A)
- Mac mini M6 32GB (id `mac-mini-m6-32`, unified): 24 GB usable, 170 GB/s, $1799 Apple list price, M6 12C CPU/12C GPU, 32GB/1TB (MHQV4LL/A)
- Mac mini M5 Pro 64GB (id `mac-mini-m5-pro-64`, unified): 48 GB usable, 307 GB/s, $3199 Apple list price, M5 Pro 18C CPU/20C GPU, 64GB/1TB (MHQU4LL/A)
- MacBook Pro M4 Max 128GB (id `macbook-pro-m4-max-128`, unified): 96 GB usable, 546 GB/s, $4999 reseller listing, 16-inch 128GB/1TB (AppleInsider price guide; sold out at listed reseller)
- MacBook Pro M5 Max 128GB (id `macbook-pro-m5-max-128`, unified): 96 GB usable, 614 GB/s, $6699 Apple list price, 14-inch M5 Max 18C CPU/40C GPU, 128GB/2TB
- MacBook Pro M6 32GB (id `macbook-pro-m6-32`, unified): 24 GB usable, 170 GB/s, $2399 Apple list price, 14-inch M6 12C CPU/12C GPU, 32GB/1TB (MHWV4LL/A; available starting Nov 6, 2026)
- Mac Studio M5 Ultra 256GB (id `mac-studio-m5-ultra-256`, unified): 192 GB usable, 1200 GB/s, $9499 Apple list price, M5 Ultra 30C CPU/64C GPU, 256GB/1TB
- Mac Studio M5 Ultra 512GB (id `mac-studio-m5-ultra-512`, unified): 384 GB usable, 1200 GB/s, $22299 Apple list price, M5 Ultra 36C CPU/80C GPU, 512GB/1TB
- Framework Desktop (Ryzen AI Max+ 395) 128GB (id `framework-desktop-128`, unified): 96 GB usable, 256 GB/s, $3449 Framework list price, DIY Edition 128GB (pre-order, ships within 5 business days)
- NVIDIA DGX Spark 64GB (id `dgx-spark-64`, unified): 48 GB usable, 273 GB/s, $4999 NVIDIA OEM-partner starting price (Acer/ASUS/Dell/Gigabyte/HP/MSI; not sold on NVIDIA store)
- NVIDIA DGX Spark 128GB (id `dgx-spark-128`, unified): 96 GB usable, 273 GB/s, $6950 NVIDIA list price (not re-verified 2026-11-05: NVIDIA marketplace blocked/empty response; last confirmed $6,950 on marketplace listing; widely reported sold out in the US)
- GeForce RTX 2060 6GB (id `rtx-2060`, gpu): 6 GB usable, 336 GB/s, $349 list (launch MSRP; discontinued)
- GeForce RTX 2060 12GB (id `rtx-2060-12gb`, gpu): 12 GB usable, 336 GB/s, $null null
- GeForce RTX 2060 Super (id `rtx-2060-super`, gpu): 8 GB usable, 448 GB/s, $399 list (launch MSRP; discontinued)
- GeForce RTX 2070 (id `rtx-2070`, gpu): 8 GB usable, 448 GB/s, $499 list (launch MSRP; discontinued)
- GeForce RTX 2070 Super (id `rtx-2070-super`, gpu): 8 GB usable, 448 GB/s, $499 list (launch MSRP; discontinued)
- GeForce RTX 2080 (id `rtx-2080`, gpu): 8 GB usable, 448 GB/s, $699 list (launch MSRP; discontinued)
- GeForce RTX 2080 Super (id `rtx-2080-super`, gpu): 8 GB usable, 496 GB/s, $699 list (launch MSRP; discontinued)
- GeForce RTX 2080 Ti (id `rtx-2080-ti`, gpu): 11 GB usable, 616 GB/s, $999 list (launch MSRP; discontinued)
- NVIDIA Titan RTX (id `titan-rtx`, gpu): 24 GB usable, 672 GB/s, $2499 list (launch MSRP; discontinued)
- GeForce RTX 3070 Ti (id `rtx-3070-ti`, gpu): 8 GB usable, 608 GB/s, $599 list (launch MSRP; discontinued)
- GeForce RTX 3080 12GB (id `rtx-3080-12gb`, gpu): 12 GB usable, 912 GB/s, $799 list (launch MSRP; discontinued)
- GeForce RTX 4060 Ti 8GB (id `rtx-4060-ti-8gb`, gpu): 8 GB usable, 288 GB/s, $399 list (launch MSRP; discontinued)
- GeForce RTX 4070 Ti (id `rtx-4070-ti`, gpu): 12 GB usable, 504 GB/s, $799 list (launch MSRP; discontinued)
- GeForce RTX 4080 (id `rtx-4080`, gpu): 16 GB usable, 717 GB/s, $1199 list (launch MSRP; discontinued)
- GeForce RTX 5060 (id `rtx-5060`, gpu): 8 GB usable, 448 GB/s, $299 list
- GeForce RTX 5060 Ti 8GB (id `rtx-5060-ti-8gb`, gpu): 8 GB usable, 448 GB/s, $379 list
- GeForce RTX 5060 Ti 16GB (id `rtx-5060-ti-16gb`, gpu): 16 GB usable, 448 GB/s, $429 list
- GeForce RTX 5070 Ti (id `rtx-5070-ti`, gpu): 16 GB usable, 896 GB/s, $749 list
- Radeon RX 6700 XT (id `rx-6700-xt`, gpu): 12 GB usable, 384 GB/s, $479 list (launch MSRP; discontinued)
- Radeon RX 6750 XT (id `rx-6750-xt`, gpu): 12 GB usable, 432 GB/s, $549 list (launch MSRP; discontinued)
- Radeon RX 6800 (id `rx-6800`, gpu): 16 GB usable, 512 GB/s, $579 list (launch MSRP; discontinued)
- Radeon RX 6800 XT (id `rx-6800-xt`, gpu): 16 GB usable, 512 GB/s, $649 list (launch MSRP; discontinued)
- Radeon RX 6900 XT (id `rx-6900-xt`, gpu): 16 GB usable, 512 GB/s, $999 list (launch MSRP; discontinued)
- Radeon RX 6950 XT (id `rx-6950-xt`, gpu): 16 GB usable, 576 GB/s, $1099 list (launch MSRP; discontinued)
- Radeon RX 7600 (id `rx-7600`, gpu): 8 GB usable, 288 GB/s, $269 list
- Radeon RX 7700 XT (id `rx-7700-xt`, gpu): 12 GB usable, 432 GB/s, $449 list
- Radeon RX 7900 GRE (id `rx-7900-gre`, gpu): 16 GB usable, 576 GB/s, $549 list
- Radeon RX 9060 XT 8GB (id `rx-9060-xt-8gb`, gpu): 8 GB usable, 320 GB/s, $299 list
- Radeon RX 9060 XT 16GB (id `rx-9060-xt-16gb`, gpu): 16 GB usable, 320 GB/s, $349 list
- Radeon RX 9070 (id `rx-9070`, gpu): 16 GB usable, 640 GB/s, $549 list
- Radeon RX 9070 XT (id `rx-9070-xt`, gpu): 16 GB usable, 640 GB/s, $599 list
- Intel Arc A750 (id `arc-a750`, gpu): 8 GB usable, 512 GB/s, $289 list (launch MSRP; discontinued)
- Intel Arc A770 16GB (id `arc-a770-16gb`, gpu): 16 GB usable, 560 GB/s, $349 list (launch MSRP; discontinued)
- Intel Arc B570 (id `arc-b570`, gpu): 10 GB usable, 380 GB/s, $219 list
- NVIDIA RTX A4000 (id `rtx-a4000`, gpu): 16 GB usable, 448 GB/s, $1000 list (launch MSRP; discontinued)
- NVIDIA RTX A5000 (id `rtx-a5000`, gpu): 24 GB usable, 768 GB/s, $2250 list (launch MSRP; discontinued)
- NVIDIA RTX A6000 (id `rtx-a6000`, gpu): 48 GB usable, 768 GB/s, $4649 list (launch MSRP; discontinued)
- NVIDIA RTX 2000 Ada 16GB (id `rtx-2000-ada`, gpu): 16 GB usable, 224 GB/s, $649 list
- NVIDIA RTX 4000 Ada 20GB (id `rtx-4000-ada`, gpu): 20 GB usable, 360 GB/s, $1250 list
- NVIDIA RTX 6000 Ada 48GB (id `rtx-6000-ada`, gpu): 48 GB usable, 960 GB/s, $6799 list
- NVIDIA RTX Pro 6000 Blackwell 96GB (id `rtx-pro-6000-blackwell`, gpu): 96 GB usable, 1792 GB/s, $8565 list
- NVIDIA Tesla P40 24GB (id `tesla-p40`, gpu): 24 GB usable, 346 GB/s, $null null
- NVIDIA Tesla V100 32GB (PCIe) (id `tesla-v100-32gb`, gpu): 32 GB usable, 900 GB/s, $null null
- AMD Radeon Instinct MI50 32GB (id `instinct-mi50-32gb`, gpu): 32 GB usable, 1024 GB/s, $null null
- AMD Radeon Instinct MI60 32GB (id `instinct-mi60`, gpu): 32 GB usable, 1024 GB/s, $null null