# MoE vs dense LLMs: VRAM and speed for local AI (2026) > Do MoE models like Qwen3-30B-A3B and gpt-oss need less VRAM? Total vs active params, memory and est. tok/s for 40 MoE models, with the cheapest GPU and Mac for each. Source: https://whatcanirunai.vercel.app/moe-vs-dense/ # MoE vs dense models for local LLMs Mixture-of-experts models like Qwen3-30B-A3B, gpt-oss and Gemma 4 26B A4B need memory for every parameter but only read a few billion per token. That makes them the best speed per GB of memory you can run locally. This page lists the 40 MoE models we track, their total and active size, the cheapest hardware for each, and side-by-side speed against similar dense models. Memory total params all experts must fit Speed active params bytes read per token MoE models tracked 40 Short answer: size your memory for the **total** parameters and expect speed close to a dense model of the **active** size. On the GeForce RTX 5090, Qwen3-30B-A3B has a speed ceiling of up to 837 tok/s against 54–77 for the dense Qwen3-32B, with about the same memory. Real MoE speeds land well below that ceiling, since routing and attention add work, but the gap stays large. ## Every MoE model we track Active parameters come from the model name (like 30B-A3B) or the maker's model card. Memory is Q4\_K\_M with \~4k context ([our formula](https://whatcanirunai.vercel.app/method/)). Speed is an upper bound on decode at Q4 on the cheapest current device that fits ([how we estimate](https://whatcanirunai.vercel.app/method/#speed)). __MoE models: total and active parameters, Q4 memory, and the cheapest GPU and Mac or mini PC that run each__ | Model | Total | Active | Active share | Q4 memory | Cheapest GPU | Cheapest Mac / mini PC | | ------------------------------------------------------------------------------------------------------- | ------- | ------ | ------------ | --------- | ----------------------------------------------------------------------------------------------- | ------------------------------------------------------------------------------------------------------ | | [Mellum2.1-12B-A2.5B-Thinking](https://whatcanirunai.vercel.app/models/mellum2.1-12b-a2.5b-thinking/)JetBrains | 12.1B | 2.5B | 21% | 9.5 GB | [Intel Arc B580](https://whatcanirunai.vercel.app/hardware/arc-b580/)$249 · up to 256 tok/s | [Mac mini M6 16GB](https://whatcanirunai.vercel.app/hardware/mac-mini-m6-16/)$899 · up to 81 tok/s | | [gpt-oss-20b](https://whatcanirunai.vercel.app/models/gpt-oss-20b/)openai | 20.9B | 3.6B | 17% | 15.6 GB | [Radeon RX 7900 XT](https://whatcanirunai.vercel.app/hardware/rx-7900-xt/)$899 · up to 312 tok/s | [Mac mini M6 24GB](https://whatcanirunai.vercel.app/hardware/mac-mini-m6-24/)$1,299 · up to 62 tok/s | | [LFM2.5-24B-A2B](https://whatcanirunai.vercel.app/models/lfm2.5-24b-a2b/)LiquidAI | 24.0B | 2.0B | 8% | 17.7 GB | [Radeon RX 7900 XT](https://whatcanirunai.vercel.app/hardware/rx-7900-xt/)$899 · up to 561 tok/s | [Mac mini M6 32GB](https://whatcanirunai.vercel.app/hardware/mac-mini-m6-32/)$1,799 · up to 112 tok/s | | [Gemma 4 26B A4B](https://whatcanirunai.vercel.app/models/gemma-4-26b-a4b-it/)google | 25.8B | 3.8B | 15% | 19.0 GB | [Intel Arc Pro B70 32GB](https://whatcanirunai.vercel.app/hardware/arc-pro-b70/)$949 · up to 224 tok/s | [Mac mini M6 32GB](https://whatcanirunai.vercel.app/hardware/mac-mini-m6-32/)$1,799 · up to 59 tok/s | | [diffusiongemma-26B-A4B-it](https://whatcanirunai.vercel.app/models/diffusiongemma-26b-a4b-it/)google | 25.8B | 4.0B | 15% | 19.0 GB | [Intel Arc Pro B70 32GB](https://whatcanirunai.vercel.app/hardware/arc-pro-b70/)$949 · up to 213 tok/s | [Mac mini M6 32GB](https://whatcanirunai.vercel.app/hardware/mac-mini-m6-32/)$1,799 · up to 56 tok/s | | [Raon-1 mini preview](https://whatcanirunai.vercel.app/models/raon-1-mini-preview/)KRAFTON | 28.8B | 4.3B | 15% | 21.1 GB | [Intel Arc Pro B70 32GB](https://whatcanirunai.vercel.app/hardware/arc-pro-b70/)$949 · up to 198 tok/s | [Mac mini M6 32GB](https://whatcanirunai.vercel.app/hardware/mac-mini-m6-32/)$1,799 · up to 52 tok/s | | [Hy-Memory-30B-A3B](https://whatcanirunai.vercel.app/models/hy-memory-30b-a3b/)tencent | 30.1B | 3.0B | 10% | 22.0 GB | [Intel Arc Pro B70 32GB](https://whatcanirunai.vercel.app/hardware/arc-pro-b70/)$949 · up to 284 tok/s | [Mac mini M5 Pro 64GB](https://whatcanirunai.vercel.app/hardware/mac-mini-m5-pro-64/)$3,199 · up to 135 tok/s | | [Qwen3-30B-A3B](https://whatcanirunai.vercel.app/models/qwen3-30b-a3b/)Qwen | 30.5B | 3.0B | 10% | 22.3 GB | [Intel Arc Pro B70 32GB](https://whatcanirunai.vercel.app/hardware/arc-pro-b70/)$949 · up to 284 tok/s | [Mac mini M5 Pro 64GB](https://whatcanirunai.vercel.app/hardware/mac-mini-m5-pro-64/)$3,199 · up to 135 tok/s | | [Qwen3-Coder-30B-A3B-Instruct](https://whatcanirunai.vercel.app/models/qwen3-coder-30b-a3b-instruct/)Qwen | 30.5B | 3.0B | 10% | 22.3 GB | [Intel Arc Pro B70 32GB](https://whatcanirunai.vercel.app/hardware/arc-pro-b70/)$949 · up to 284 tok/s | [Mac mini M5 Pro 64GB](https://whatcanirunai.vercel.app/hardware/mac-mini-m5-pro-64/)$3,199 · up to 135 tok/s | | [Qwen3-30B-A3B-Instruct-2507](https://whatcanirunai.vercel.app/models/qwen3-30b-a3b-instruct-2507/)Qwen | 30.5B | 3.0B | 10% | 22.3 GB | [Intel Arc Pro B70 32GB](https://whatcanirunai.vercel.app/hardware/arc-pro-b70/)$949 · up to 284 tok/s | [Mac mini M5 Pro 64GB](https://whatcanirunai.vercel.app/hardware/mac-mini-m5-pro-64/)$3,199 · up to 135 tok/s | | [GLM-4.7-Flash](https://whatcanirunai.vercel.app/models/glm-4.7-flash/)zai-org | 31.2B | 3.0B | 10% | 22.8 GB | [Intel Arc Pro B70 32GB](https://whatcanirunai.vercel.app/hardware/arc-pro-b70/)$949 · up to 284 tok/s | [Mac mini M5 Pro 64GB](https://whatcanirunai.vercel.app/hardware/mac-mini-m5-pro-64/)$3,199 · up to 135 tok/s | | [NVIDIA-Nemotron-3.5-Lightning-30B-A3B-BF16](https://whatcanirunai.vercel.app/models/nvidia-nemotron-3.5-lightning-30b-a3b-bf16/)nvidia | 31.6B | 3.0B | 10% | 23.0 GB | [Intel Arc Pro B70 32GB](https://whatcanirunai.vercel.app/hardware/arc-pro-b70/)$949 · up to 284 tok/s | [Mac mini M5 Pro 64GB](https://whatcanirunai.vercel.app/hardware/mac-mini-m5-pro-64/)$3,199 · up to 135 tok/s | | [NVIDIA-Nemotron-3-Nano-30B-A3B-BF16](https://whatcanirunai.vercel.app/models/nvidia-nemotron-3-nano-30b-a3b-bf16/)nvidia | 31.6B | 3.0B | 10% | 23.0 GB | [Intel Arc Pro B70 32GB](https://whatcanirunai.vercel.app/hardware/arc-pro-b70/)$949 · up to 284 tok/s | [Mac mini M5 Pro 64GB](https://whatcanirunai.vercel.app/hardware/mac-mini-m5-pro-64/)$3,199 · up to 135 tok/s | | [MemT-1.0-Nano](https://whatcanirunai.vercel.app/models/memt-1.0-nano/)MemTensor | 35.1B | 3.0B | 9% | 25.5 GB | [Intel Arc Pro B70 32GB](https://whatcanirunai.vercel.app/hardware/arc-pro-b70/)$949 · up to 284 tok/s | [Mac mini M5 Pro 64GB](https://whatcanirunai.vercel.app/hardware/mac-mini-m5-pro-64/)$3,199 · up to 135 tok/s | | [Ornith-1.5-35B-A3B](https://whatcanirunai.vercel.app/models/ornith-1.5-35b-a3b/)ornith-ai | 36.0B | 3.0B | 8% | 26.1 GB | [Intel Arc Pro B70 32GB](https://whatcanirunai.vercel.app/hardware/arc-pro-b70/)$949 · up to 284 tok/s | [Mac mini M5 Pro 64GB](https://whatcanirunai.vercel.app/hardware/mac-mini-m5-pro-64/)$3,199 · up to 135 tok/s | | [Qwen3.6-35B-A3B](https://whatcanirunai.vercel.app/models/qwen3.6-35b-a3b/)Qwen | 36.0B | 3.0B | 8% | 26.1 GB | [Intel Arc Pro B70 32GB](https://whatcanirunai.vercel.app/hardware/arc-pro-b70/)$949 · up to 284 tok/s | [Mac mini M5 Pro 64GB](https://whatcanirunai.vercel.app/hardware/mac-mini-m5-pro-64/)$3,199 · up to 135 tok/s | | [Qwen3.5-35B-A3B](https://whatcanirunai.vercel.app/models/qwen3.5-35b-a3b/)Qwen | 36.0B | 3.0B | 8% | 26.1 GB | [Intel Arc Pro B70 32GB](https://whatcanirunai.vercel.app/hardware/arc-pro-b70/)$949 · up to 284 tok/s | [Mac mini M5 Pro 64GB](https://whatcanirunai.vercel.app/hardware/mac-mini-m5-pro-64/)$3,199 · up to 135 tok/s | | [Recursive Flint](https://whatcanirunai.vercel.app/models/flint/)RecursiveSI | 39.8B | 2.5B | 6% | 28.7 GB | [Intel Arc Pro B70 32GB](https://whatcanirunai.vercel.app/hardware/arc-pro-b70/)$949 · up to 341 tok/s | [Mac mini M5 Pro 64GB](https://whatcanirunai.vercel.app/hardware/mac-mini-m5-pro-64/)$3,199 · up to 162 tok/s | | [gpt-oss-120b](https://whatcanirunai.vercel.app/models/gpt-oss-120b/)openai | 116.8B | 5.1B | 4% | 82.5 GB | [NVIDIA RTX Pro 6000 Blackwell 96GB](https://whatcanirunai.vercel.app/hardware/rtx-pro-6000-blackwell/)$8,565 · up to 493 tok/s | [Framework Desktop (Ryzen AI Max+ 395) 128GB](https://whatcanirunai.vercel.app/hardware/framework-desktop-128/)$3,449 · up to 66 tok/s | | [Ling-3.0-flash-Sante](https://whatcanirunai.vercel.app/models/ling-3.0-flash-sante/)inclusionAI | 127.5B | 5.1B | 4% | 89.9 GB | none we track | [Mac Studio M5 Ultra 256GB](https://whatcanirunai.vercel.app/hardware/mac-studio-m5-ultra-256/)$9,499 · up to 310 tok/s | | [MiniMax-M2.5](https://whatcanirunai.vercel.app/models/minimax-m2.5/)MiniMaxAI | 228.7B | 10.0B | 4% | 160.4 GB | none we track | [Mac Studio M5 Ultra 256GB](https://whatcanirunai.vercel.app/hardware/mac-studio-m5-ultra-256/)$9,499 · up to 158 tok/s | | [MemT-1.0-Flash](https://whatcanirunai.vercel.app/models/memt-1.0-flash/)MemTensor | 287.9B | 17.0B | 6% | 201.8 GB | none we track | [Mac Studio M5 Ultra 512GB](https://whatcanirunai.vercel.app/hardware/mac-studio-m5-ultra-512/)$22,299 · up to 93 tok/s | | [DeepSeek-V4-Flash](https://whatcanirunai.vercel.app/models/deepseek-v4-flash/)deepseek-ai | 290.9B | 13.0B | 4% | 203.8 GB | none we track | [Mac Studio M5 Ultra 512GB](https://whatcanirunai.vercel.app/hardware/mac-studio-m5-ultra-512/)$22,299 · up to 122 tok/s | | [Hy3](https://whatcanirunai.vercel.app/models/hy3/)tencent | 298.8B | 21.0B | 7% | 209.3 GB | none we track | [Mac Studio M5 Ultra 512GB](https://whatcanirunai.vercel.app/hardware/mac-studio-m5-ultra-512/)$22,299 · up to 75 tok/s | | [DeepSeek-V4-Flash-0731](https://whatcanirunai.vercel.app/models/deepseek-v4-flash-0731/)deepseek-ai | 304.2B | 13.0B | 4% | 213.1 GB | none we track | [Mac Studio M5 Ultra 512GB](https://whatcanirunai.vercel.app/hardware/mac-studio-m5-ultra-512/)$22,299 · up to 122 tok/s | | [GLM-5.3-Flash](https://whatcanirunai.vercel.app/models/glm-5.3-flash/)zai-org | 321.3B | 18.0B | 6% | 225.0 GB | none we track | [Mac Studio M5 Ultra 512GB](https://whatcanirunai.vercel.app/hardware/mac-studio-m5-ultra-512/)$22,299 · up to 88 tok/s | | [Qwen3.5-397B-A17B](https://whatcanirunai.vercel.app/models/qwen3.5-397b-a17b/)Qwen | 403.4B | 17.0B | 4% | 282.2 GB | none we track | [Mac Studio M5 Ultra 512GB](https://whatcanirunai.vercel.app/hardware/mac-studio-m5-ultra-512/)$22,299 · up to 93 tok/s | | [Beam-1.0-501B-A23B](https://whatcanirunai.vercel.app/models/beam-1.0-501b-a23b/)reflection | 501.3B | 23.0B | 5% | 350.5 GB | none we track | none we track | | [DeepSeek-R1](https://whatcanirunai.vercel.app/models/deepseek-r1/)deepseek-ai | 684.5B | 37.0B | 5% | 478.2 GB | none we track | none we track | | [DeepSeek-V3-0324](https://whatcanirunai.vercel.app/models/deepseek-v3-0324/)deepseek-ai | 684.5B | 37.0B | 5% | 478.2 GB | none we track | none we track | | [DeepSeek-V3.2](https://whatcanirunai.vercel.app/models/deepseek-v3.2/)deepseek-ai | 685.4B | 37.0B | 5% | 478.8 GB | none we track | none we track | | [GLM-5.2](https://whatcanirunai.vercel.app/models/glm-5.2/)zai-org | 753.3B | 40.0B | 5% | 526.2 GB | none we track | none we track | | [GLM-5](https://whatcanirunai.vercel.app/models/glm-5/)zai-org | 753.9B | 40.0B | 5% | 526.6 GB | none we track | none we track | | [GLM-5.1](https://whatcanirunai.vercel.app/models/glm-5.1/)zai-org | 753.9B | 40.0B | 5% | 526.6 GB | none we track | none we track | | [Muse-Spark-Open-816B-A42B](https://whatcanirunai.vercel.app/models/muse-spark-open-816b-a42b/)meta-models | 822.0B | 42.0B | 5% | 574.1 GB | none we track | none we track | | [Inkling](https://whatcanirunai.vercel.app/models/inkling/)thinkingmachines | 952.4B | 41.0B | 4% | 665.0 GB | none we track | none we track | | [Kimi-K2-Instruct](https://whatcanirunai.vercel.app/models/kimi-k2-instruct/)moonshotai | 1026.4B | 32.0B | 3% | 716.6 GB | none we track | none we track | | [Kimi-K2.5](https://whatcanirunai.vercel.app/models/kimi-k2.5/)moonshotai | 1026.9B | 32.0B | 3% | 716.9 GB | none we track | none we track | | [DeepSeek-V4-Pro](https://whatcanirunai.vercel.app/models/deepseek-v4-pro/)deepseek-ai | 1598.8B | 49.0B | 3% | 1115.7 GB | none we track | none we track | | [Kimi-K3](https://whatcanirunai.vercel.app/models/kimi-k3/)moonshotai | 2779.9B | 104.0B | 4% | 1939.1 GB | none we track | none we track | Showing the 40 most popular of 226 MoE models. [See all 226 in the model directory →](https://whatcanirunai.vercel.app/models/?moe=1) Models whose makers don't state active parameters (for example GLM-5.3 and Sage-A1.5) are left out here; on their own pages we estimate speed from total parameters, which understates it. ## MoE vs dense, head to head Pairs of similar memory size on 7 current devices. MoE figures are upper bounds; dense figures are a range. ### Qwen3-30B-A3B vs Qwen3-32B [Qwen3-30B-A3B](https://whatcanirunai.vercel.app/models/qwen3-30b-a3b/): 30.5B total, 3.0B active, \~22.3 GB at Q4\. [Qwen3-32B](https://whatcanirunai.vercel.app/models/qwen3-32b/): 32.8B dense, \~23.8 GB at Q4\. Est. decode tok/s at Q4. __Estimated tokens per second, Qwen3-30B-A3B vs Qwen3-32B__ | Device | Qwen3-30B-A3B (MoE) | Qwen3-32B (dense) | | --------------------------------------------------------------------------------------------------------------- | ------------------- | ----------------- | | [GeForce RTX 4060 Ti 16GB](https://whatcanirunai.vercel.app/hardware/rtx-4060-ti-16gb/)16 GB usable · 288 GB/s · $499 | doesn’t fit | doesn’t fit | | [Radeon RX 7900 XTX](https://whatcanirunai.vercel.app/hardware/rx-7900-xtx/)24 GB usable · 960 GB/s · $999 | doesn’t fit | doesn’t fit | | [GeForce RTX 5090](https://whatcanirunai.vercel.app/hardware/rtx-5090/)32 GB usable · 1792 GB/s · $1,999 | up to 837 | 54–77 | | [Mac mini M6 32GB](https://whatcanirunai.vercel.app/hardware/mac-mini-m6-32/)24 GB usable · 170 GB/s · $1,799 | doesn’t fit | doesn’t fit | | [Mac mini M5 Pro 64GB](https://whatcanirunai.vercel.app/hardware/mac-mini-m5-pro-64/)48 GB usable · 307 GB/s · $3,199 | up to 135 | 7–12 | | [Framework Desktop (Ryzen AI Max+ 395) 128GB](https://whatcanirunai.vercel.app/hardware/framework-desktop-128/)96 GB usable · 256 GB/s · $3,449 | up to 113 | 6–10 | | [MacBook Pro M5 Max 128GB](https://whatcanirunai.vercel.app/hardware/macbook-pro-m5-max-128/)96 GB usable · 614 GB/s · $6,699 | up to 270 | 14–25 | ### Gemma 4 26B A4B vs Gemma 4 31B [Gemma 4 26B A4B](https://whatcanirunai.vercel.app/models/gemma-4-26b-a4b-it/): 25.8B total, 3.8B active, \~19.0 GB at Q4\. [Gemma 4 31B](https://whatcanirunai.vercel.app/models/gemma-4-31b-it/): 31.3B dense, \~22.8 GB at Q4\. Est. decode tok/s at Q4. __Estimated tokens per second, Gemma 4 26B A4B vs Gemma 4 31B__ | Device | Gemma 4 26B A4B (MoE) | Gemma 4 31B (dense) | | --------------------------------------------------------------------------------------------------------------- | --------------------- | ------------------- | | [GeForce RTX 4060 Ti 16GB](https://whatcanirunai.vercel.app/hardware/rtx-4060-ti-16gb/)16 GB usable · 288 GB/s · $499 | doesn’t fit | doesn’t fit | | [Radeon RX 7900 XTX](https://whatcanirunai.vercel.app/hardware/rx-7900-xtx/)24 GB usable · 960 GB/s · $999 | up to 354 | doesn’t fit | | [GeForce RTX 5090](https://whatcanirunai.vercel.app/hardware/rtx-5090/)32 GB usable · 1792 GB/s · $1,999 | up to 661 | 57–80 | | [Mac mini M6 32GB](https://whatcanirunai.vercel.app/hardware/mac-mini-m6-32/)24 GB usable · 170 GB/s · $1,799 | up to 59 | doesn’t fit | | [Mac mini M5 Pro 64GB](https://whatcanirunai.vercel.app/hardware/mac-mini-m5-pro-64/)48 GB usable · 307 GB/s · $3,199 | up to 107 | 7–13 | | [Framework Desktop (Ryzen AI Max+ 395) 128GB](https://whatcanirunai.vercel.app/hardware/framework-desktop-128/)96 GB usable · 256 GB/s · $3,449 | up to 89 | 6–11 | | [MacBook Pro M5 Max 128GB](https://whatcanirunai.vercel.app/hardware/macbook-pro-m5-max-128/)96 GB usable · 614 GB/s · $6,699 | up to 213 | 15–26 | ### gpt-oss-120b vs Llama-3.3-70B-Instruct [gpt-oss-120b](https://whatcanirunai.vercel.app/models/gpt-oss-120b/): 116.8B total, 5.1B active, \~82.5 GB at Q4\. [Llama-3.3-70B-Instruct](https://whatcanirunai.vercel.app/models/llama-3.3-70b-instruct/): 70.6B dense, \~50.2 GB at Q4\. Est. decode tok/s at Q4. __Estimated tokens per second, gpt-oss-120b vs Llama-3.3-70B-Instruct__ | Device | gpt-oss-120b (MoE) | Llama-3.3-70B-Instruct (dense) | | --------------------------------------------------------------------------------------------------------------- | ------------------ | ------------------------------ | | [GeForce RTX 4060 Ti 16GB](https://whatcanirunai.vercel.app/hardware/rtx-4060-ti-16gb/)16 GB usable · 288 GB/s · $499 | doesn’t fit | doesn’t fit | | [Radeon RX 7900 XTX](https://whatcanirunai.vercel.app/hardware/rx-7900-xtx/)24 GB usable · 960 GB/s · $999 | doesn’t fit | doesn’t fit | | [GeForce RTX 5090](https://whatcanirunai.vercel.app/hardware/rtx-5090/)32 GB usable · 1792 GB/s · $1,999 | doesn’t fit | doesn’t fit | | [Mac mini M6 32GB](https://whatcanirunai.vercel.app/hardware/mac-mini-m6-32/)24 GB usable · 170 GB/s · $1,799 | doesn’t fit | doesn’t fit | | [Mac mini M5 Pro 64GB](https://whatcanirunai.vercel.app/hardware/mac-mini-m5-pro-64/)48 GB usable · 307 GB/s · $3,199 | doesn’t fit | doesn’t fit | | [Framework Desktop (Ryzen AI Max+ 395) 128GB](https://whatcanirunai.vercel.app/hardware/framework-desktop-128/)96 GB usable · 256 GB/s · $3,449 | up to 66 | 3–5 | | [MacBook Pro M5 Max 128GB](https://whatcanirunai.vercel.app/hardware/macbook-pro-m5-max-128/)96 GB usable · 614 GB/s · $6,699 | up to 159 | 6–11 | ### gpt-oss-20b vs Qwen3-14B [gpt-oss-20b](https://whatcanirunai.vercel.app/models/gpt-oss-20b/): 20.9B total, 3.6B active, \~15.6 GB at Q4\. [Qwen3-14B](https://whatcanirunai.vercel.app/models/qwen3-14b/): 14.8B dense, \~11.3 GB at Q4\. Est. decode tok/s at Q4. __Estimated tokens per second, gpt-oss-20b vs Qwen3-14B__ | Device | gpt-oss-20b (MoE) | Qwen3-14B (dense) | | --------------------------------------------------------------------------------------------------------------- | ----------------- | ----------------- | | [GeForce RTX 4060 Ti 16GB](https://whatcanirunai.vercel.app/hardware/rtx-4060-ti-16gb/)16 GB usable · 288 GB/s · $499 | doesn’t fit | 19–27 | | [Radeon RX 7900 XTX](https://whatcanirunai.vercel.app/hardware/rx-7900-xtx/)24 GB usable · 960 GB/s · $999 | up to 374 | 64–91 | | [GeForce RTX 5090](https://whatcanirunai.vercel.app/hardware/rtx-5090/)32 GB usable · 1792 GB/s · $1,999 | up to 698 | 120–170 | | [Mac mini M6 32GB](https://whatcanirunai.vercel.app/hardware/mac-mini-m6-32/)24 GB usable · 170 GB/s · $1,799 | up to 62 | 9–15 | | [Mac mini M5 Pro 64GB](https://whatcanirunai.vercel.app/hardware/mac-mini-m5-pro-64/)48 GB usable · 307 GB/s · $3,199 | up to 113 | 15–27 | | [Framework Desktop (Ryzen AI Max+ 395) 128GB](https://whatcanirunai.vercel.app/hardware/framework-desktop-128/)96 GB usable · 256 GB/s · $3,449 | up to 94 | 13–23 | | [MacBook Pro M5 Max 128GB](https://whatcanirunai.vercel.app/hardware/macbook-pro-m5-max-128/)96 GB usable · 614 GB/s · $6,699 | up to 225 | 31–55 | ## Quick answers What is a mixture-of-experts (MoE) model? A MoE model splits most of its layers into many small "experts" and a router picks a few of them for each token. So it has a large total parameter count, but only a fraction (the active parameters) does work on each token. Qwen3-30B-A3B, for example, has 30.5B parameters in total but uses about 3.0B per token. Does a MoE model need less VRAM? No. All the experts must be in memory, because any of them can be picked for the next token. Memory follows total parameters, just like a dense model of the same size. Qwen3-30B-A3B needs \~22.3 GB at Q4, close to Qwen3-32B's \~23.8 GB. Why are MoE models so much faster locally? Generating a token is limited by how many bytes are read from memory, and a MoE model reads only its active experts. With 10% of weights active, decode speed can be several times a dense model of the same size. We show MoE speed as an upper bound ("up to") because routing, attention and shared layers add cost we haven't calibrated. Is MoE a good fit for Macs and unified-memory PCs? Yes. Macs, AMD Ryzen AI Max machines and DGX Spark have lots of memory but less bandwidth than a high-end GPU, so dense 70B models run slowly on them. A big MoE model fits in that memory and still runs at a usable speed. On the Framework Desktop (Ryzen AI Max+ 395) 128GB, our estimate for gpt-oss-120b is up to 66 tok/s, against 3–5 tok/s for Llama-3.3-70B-Instruct. Is a MoE model as smart as a dense model of the same total size? Usually it lands between a dense model of its active size and one of its total size, and it varies by model. Check benchmarks for your task; this page covers memory and speed only. Can I offload MoE experts to system RAM? Yes. llama.cpp can keep the experts in system RAM and the rest on the GPU (for example with --n-cpu-moe or tensor overrides), which lets a big MoE model run on a smaller GPU at a lower speed. We don't model offloading; our fit figures assume the whole model is in GPU or unified memory. [Tokens per second](https://whatcanirunai.vercel.app/tokens-per-second/) · [Q4 vs Q8 vs FP16](https://whatcanirunai.vercel.app/quantization/) · [Context length and VRAM](https://whatcanirunai.vercel.app/context-length/) · [NVIDIA vs AMD vs Mac](https://whatcanirunai.vercel.app/nvidia-vs-amd-vs-mac/) · [VRAM calculator](https://whatcanirunai.vercel.app/vram-calculator/) --- Agent guide: https://whatcanirunai.vercel.app/llms.txt · JSON API: https://whatcanirunai.vercel.app/developers/