# Ollama VRAM requirements: GPU memory for every model (2026) > How much VRAM each Ollama model needs: real download sizes and quants for 39 tags (Llama, Qwen3, Gemma, gpt-oss, DeepSeek R1), est. memory at 4k/32k context, cheapest GPU and Mac. Source: https://whatcanirunai.vercel.app/ollama-vram-requirements/ # Ollama VRAM requirements: how much GPU memory each model needs How much VRAM a model needs in Ollama, tag by tag. Each row uses the real model file size, quantization and context window from Ollama's registry for 39 popular tags, an estimate of total memory at 4k and 32k context, and the cheapest GPU and Mac we track that runs it fully on the GPU. Tags 39 Ollama library defaults Default quant Q4\_K\_M for most tags File sizes 2026-11-12 from registry.ollama.ai Short answer: a Q4 Ollama model needs about its model file size plus 15% plus 1 GB. That means \~6.7 GB for an 8B tag, \~16.9 GB for gpt-oss:20b, \~24.2 GB for a 32B tag and \~49.9 GB for a 70B tag at Ollama's default 4k context. More context needs more memory. ## How we estimate it memory ≈ Ollama model file size × 1.15 + 1 GB (≈4k context) · 32k adds the KV cache from the model's published config (FP16 cache) Same formula as the rest of the site ([how we calculate](https://whatcanirunai.vercel.app/method/)), applied to the actual model file Ollama pulls instead of a bits-per-weight guess. These are estimates, not measured with `ollama ps`. "Smallest card" is the smallest common VRAM size where the model uses 90% or less. Cheapest GPU and Mac are the cheapest current devices we track that fit at 4k, at dated list prices. Mac memory is unified, and macOS keeps some of it, so we use the usable amount. ## VRAM by Ollama tag __Ollama tag, download size and quantization, estimated memory at 4k and 32k context, smallest VRAM size, and cheapest GPU and Mac that fit__ | Ollama tag | Model file | Est. memory, 4k ctx | Est. memory, 32k ctx | Smallest card | Cheapest GPU | Cheapest Mac / mini PC | | ------------------------------------------------------------------------------------------------------- | ---------------- | ------------------- | -------------------- | ------------- | ---------------------------------------------------------------------------------- | ------------------------------------------------------------------------------------------ | | llama3.2:1b[Llama-3.2-1B-Instruct](https://whatcanirunai.vercel.app/models/llama-3.2-1b-instruct/) | 1.3 GBQ8\_0 | **2.5 GB** | 3.5 GB | 4 GB | [Intel Arc B570](https://whatcanirunai.vercel.app/hardware/arc-b570/)$219 list | [Mac mini M6 16GB](https://whatcanirunai.vercel.app/hardware/mac-mini-m6-16/)$899 list | | qwen3:1.7b[Qwen3-1.7B](https://whatcanirunai.vercel.app/models/qwen3-1.7b/) | 1.4 GBQ4\_K\_M | **2.6 GB** | 5.9 GB | 4 GB | [Intel Arc B570](https://whatcanirunai.vercel.app/hardware/arc-b570/)$219 list | [Mac mini M6 16GB](https://whatcanirunai.vercel.app/hardware/mac-mini-m6-16/)$899 list | | llama3.2:3b[Llama-3.2-3B-Instruct](https://whatcanirunai.vercel.app/models/llama-3.2-3b-instruct/) | 2.0 GBQ4\_K\_M | **3.3 GB** | 6.6 GB | 4 GB | [Intel Arc B570](https://whatcanirunai.vercel.app/hardware/arc-b570/)$219 list | [Mac mini M6 16GB](https://whatcanirunai.vercel.app/hardware/mac-mini-m6-16/)$899 list | | phi4-mini:3.8b[Phi-4-mini-instruct](https://whatcanirunai.vercel.app/models/phi-4-mini-instruct/) | 2.5 GBQ4\_K\_M | **3.9 GB** | 7.6 GB | 6 GB | [Intel Arc B570](https://whatcanirunai.vercel.app/hardware/arc-b570/)$219 list | [Mac mini M6 16GB](https://whatcanirunai.vercel.app/hardware/mac-mini-m6-16/)$899 list | | qwen3:4b[Qwen3-4B](https://whatcanirunai.vercel.app/models/qwen3-4b/) | 2.5 GBQ4\_K\_M | **3.9 GB** | 8.1 GB | 6 GB | [Intel Arc B570](https://whatcanirunai.vercel.app/hardware/arc-b570/)$219 list | [Mac mini M6 16GB](https://whatcanirunai.vercel.app/hardware/mac-mini-m6-16/)$899 list | | gemma3:4b[gemma-3-4b-it](https://whatcanirunai.vercel.app/models/gemma-3-4b-it/) | 2.5 GBQ4\_K\_M | **3.9 GB** | — | 6 GB | [Intel Arc B570](https://whatcanirunai.vercel.app/hardware/arc-b570/)$219 list | [Mac mini M6 16GB](https://whatcanirunai.vercel.app/hardware/mac-mini-m6-16/)$899 list | | gemma4:e2b[Gemma 4 E2B](https://whatcanirunai.vercel.app/models/gemma-4-e2b-it/) | 3.5 GBQ4\_K\_M | **5.0 GB** | — | 6 GB | [Intel Arc B570](https://whatcanirunai.vercel.app/hardware/arc-b570/)$219 list | [Mac mini M6 16GB](https://whatcanirunai.vercel.app/hardware/mac-mini-m6-16/)$899 list | | mistral:7b[Mistral-7B-Instruct-v0.3](https://whatcanirunai.vercel.app/models/mistral-7b-instruct-v0.3/) | 4.4 GBQ4\_K\_M | **6.0 GB** | 9.8 GB | 8 GB | [Intel Arc B570](https://whatcanirunai.vercel.app/hardware/arc-b570/)$219 list | [Mac mini M6 16GB](https://whatcanirunai.vercel.app/hardware/mac-mini-m6-16/)$899 list | | deepseek-r1:7b[DeepSeek-R1-Distill-Qwen-7B](https://whatcanirunai.vercel.app/models/deepseek-r1-distill-qwen-7b/) | 4.7 GBQ4\_K\_M | **6.4 GB** | 8.0 GB | 8 GB | [Intel Arc B570](https://whatcanirunai.vercel.app/hardware/arc-b570/)$219 list | [Mac mini M6 16GB](https://whatcanirunai.vercel.app/hardware/mac-mini-m6-16/)$899 list | | qwen2.5-coder:7b[Qwen2.5-Coder-7B-Instruct](https://whatcanirunai.vercel.app/models/qwen2.5-coder-7b-instruct/) | 4.7 GBQ4\_K\_M | **6.4 GB** | 8.0 GB | 8 GB | [Intel Arc B570](https://whatcanirunai.vercel.app/hardware/arc-b570/)$219 list | [Mac mini M6 16GB](https://whatcanirunai.vercel.app/hardware/mac-mini-m6-16/)$899 list | | llama3.1:8b[Llama-3.1-8B-Instruct](https://whatcanirunai.vercel.app/models/llama-3.1-8b-instruct/) | 4.9 GBQ4\_K\_M | **6.7 GB** | 10.4 GB | 8 GB | [Intel Arc B570](https://whatcanirunai.vercel.app/hardware/arc-b570/)$219 list | [Mac mini M6 16GB](https://whatcanirunai.vercel.app/hardware/mac-mini-m6-16/)$899 list | | granite3.3:8b[granite-3.3-8b-instruct](https://whatcanirunai.vercel.app/models/granite-3.3-8b-instruct/) | 4.9 GBQ4\_K\_M | **6.7 GB** | 11.4 GB | 8 GB | [Intel Arc B570](https://whatcanirunai.vercel.app/hardware/arc-b570/)$219 list | [Mac mini M6 16GB](https://whatcanirunai.vercel.app/hardware/mac-mini-m6-16/)$899 list | | qwen3:8b[Qwen3-8B](https://whatcanirunai.vercel.app/models/qwen3-8b/) | 5.2 GBQ4\_K\_M | **7.0 GB** | 11.2 GB | 8 GB | [Intel Arc B570](https://whatcanirunai.vercel.app/hardware/arc-b570/)$219 list | [Mac mini M6 16GB](https://whatcanirunai.vercel.app/hardware/mac-mini-m6-16/)$899 list | | deepseek-r1:8b[DeepSeek-R1-0528-Qwen3-8B](https://whatcanirunai.vercel.app/models/deepseek-r1-0528-qwen3-8b/) | 5.2 GBQ4\_K\_M | **7.0 GB** | 11.2 GB | 8 GB | [Intel Arc B570](https://whatcanirunai.vercel.app/hardware/arc-b570/)$219 list | [Mac mini M6 16GB](https://whatcanirunai.vercel.app/hardware/mac-mini-m6-16/)$899 list | | gemma4:e4b[Gemma 4 E4B](https://whatcanirunai.vercel.app/models/gemma-4-e4b-it/) | 5.5 GBQ4\_K\_M | **7.3 GB** | — | 12 GB | [Intel Arc B570](https://whatcanirunai.vercel.app/hardware/arc-b570/)$219 list | [Mac mini M6 16GB](https://whatcanirunai.vercel.app/hardware/mac-mini-m6-16/)$899 list | | mistral-nemo:12b[Mistral-Nemo-Instruct-2407](https://whatcanirunai.vercel.app/models/mistral-nemo-instruct-2407/) | 7.1 GBQ4\_0 | **9.1 GB** | 13.8 GB | 12 GB | [Intel Arc B580](https://whatcanirunai.vercel.app/hardware/arc-b580/)$249 list | [Mac mini M6 16GB](https://whatcanirunai.vercel.app/hardware/mac-mini-m6-16/)$899 list | | gemma3:12b[gemma-3-12b-it](https://whatcanirunai.vercel.app/models/gemma-3-12b-it/) | 7.3 GBQ4\_K\_M | **9.4 GB** | — | 12 GB | [Intel Arc B580](https://whatcanirunai.vercel.app/hardware/arc-b580/)$249 list | [Mac mini M6 16GB](https://whatcanirunai.vercel.app/hardware/mac-mini-m6-16/)$899 list | | gemma4:12b[Gemma 4 12B](https://whatcanirunai.vercel.app/models/gemma-4-12b-it/) | 7.4 GBQ4\_K\_M | **9.5 GB** | — | 12 GB | [Intel Arc B580](https://whatcanirunai.vercel.app/hardware/arc-b580/)$249 list | [Mac mini M6 16GB](https://whatcanirunai.vercel.app/hardware/mac-mini-m6-16/)$899 list | | olmo2:13b[OLMo-2-1124-13B-Instruct](https://whatcanirunai.vercel.app/models/olmo-2-1124-13b-instruct/) | 8.3 GBQ4\_K\_M | **10.6 GB** | max 4k ctx | 12 GB | [Intel Arc B580](https://whatcanirunai.vercel.app/hardware/arc-b580/)$249 list | [Mac mini M6 16GB](https://whatcanirunai.vercel.app/hardware/mac-mini-m6-16/)$899 list | | deepseek-r1:14b[DeepSeek-R1-Distill-Qwen-14B](https://whatcanirunai.vercel.app/models/deepseek-r1-distill-qwen-14b/) | 9.0 GBQ4\_K\_M | **11.3 GB** | 17.0 GB | 16 GB | [Radeon RX 7600 XT](https://whatcanirunai.vercel.app/hardware/rx-7600-xt/)$329 list | [Mac mini M6 24GB](https://whatcanirunai.vercel.app/hardware/mac-mini-m6-24/)$1,299 list | | qwen2.5-coder:14b[Qwen2.5-Coder-14B-Instruct](https://whatcanirunai.vercel.app/models/qwen2.5-coder-14b-instruct/) | 9.0 GBQ4\_K\_M | **11.3 GB** | 17.0 GB | 16 GB | [Radeon RX 7600 XT](https://whatcanirunai.vercel.app/hardware/rx-7600-xt/)$329 list | [Mac mini M6 24GB](https://whatcanirunai.vercel.app/hardware/mac-mini-m6-24/)$1,299 list | | phi4:14b[phi-4](https://whatcanirunai.vercel.app/models/phi-4/) | 9.1 GBQ4\_K\_M | **11.4 GB** | max 16k ctx | 16 GB | [Radeon RX 7600 XT](https://whatcanirunai.vercel.app/hardware/rx-7600-xt/)$329 list | [Mac mini M6 24GB](https://whatcanirunai.vercel.app/hardware/mac-mini-m6-24/)$1,299 list | | qwen3:14b[Qwen3-14B](https://whatcanirunai.vercel.app/models/qwen3-14b/) | 9.3 GBQ4\_K\_M | **11.7 GB** | 16.4 GB | 16 GB | [Radeon RX 7600 XT](https://whatcanirunai.vercel.app/hardware/rx-7600-xt/)$329 list | [Mac mini M6 24GB](https://whatcanirunai.vercel.app/hardware/mac-mini-m6-24/)$1,299 list | | gpt-oss:20b[gpt-oss-20b](https://whatcanirunai.vercel.app/models/gpt-oss-20b/) | 13.8 GBMXFP4 | **16.9 GB** | 17.6 GB | 24 GB | [Radeon RX 7900 XT](https://whatcanirunai.vercel.app/hardware/rx-7900-xt/)$899 list | [Mac mini M6 32GB](https://whatcanirunai.vercel.app/hardware/mac-mini-m6-32/)$1,799 list | | mistral-small3.2:24b[Mistral-Small-3.2-24B-Instruct-2506](https://whatcanirunai.vercel.app/models/mistral-small-3.2-24b-instruct-2506/) | 15.2 GBQ4\_K\_M | **18.5 GB** | 23.2 GB | 24 GB | [Intel Arc Pro B70 32GB](https://whatcanirunai.vercel.app/hardware/arc-pro-b70/)$949 list | [Mac mini M6 32GB](https://whatcanirunai.vercel.app/hardware/mac-mini-m6-32/)$1,799 list | | gemma3:27b[gemma-3-27b-it](https://whatcanirunai.vercel.app/models/gemma-3-27b-it/) | 16.6 GBQ4\_K\_M | **20.0 GB** | — | 24 GB | [Intel Arc Pro B70 32GB](https://whatcanirunai.vercel.app/hardware/arc-pro-b70/)$949 list | [Mac mini M6 32GB](https://whatcanirunai.vercel.app/hardware/mac-mini-m6-32/)$1,799 list | | gemma4:26b[Gemma 4 26B A4B](https://whatcanirunai.vercel.app/models/gemma-4-26b-a4b-it/) | 17.1 GBQ4\_K\_M | **20.6 GB** | — | 24 GB | [Intel Arc Pro B70 32GB](https://whatcanirunai.vercel.app/hardware/arc-pro-b70/)$949 list | [Mac mini M6 32GB](https://whatcanirunai.vercel.app/hardware/mac-mini-m6-32/)$1,799 list | | qwen3:30b[Qwen3-30B-A3B](https://whatcanirunai.vercel.app/models/qwen3-30b-a3b/) | 18.6 GBQ4\_K\_M | **22.3 GB** | 25.2 GB | 32 GB | [Intel Arc Pro B70 32GB](https://whatcanirunai.vercel.app/hardware/arc-pro-b70/)$949 list | [Mac mini M5 Pro 64GB](https://whatcanirunai.vercel.app/hardware/mac-mini-m5-pro-64/)$3,199 list | | qwen3-coder:30b[Qwen3-Coder-30B-A3B-Instruct](https://whatcanirunai.vercel.app/models/qwen3-coder-30b-a3b-instruct/) | 18.6 GBQ4\_K\_M | **22.3 GB** | 25.2 GB | 32 GB | [Intel Arc Pro B70 32GB](https://whatcanirunai.vercel.app/hardware/arc-pro-b70/)$949 list | [Mac mini M5 Pro 64GB](https://whatcanirunai.vercel.app/hardware/mac-mini-m5-pro-64/)$3,199 list | | gemma4:31b[Gemma 4 31B](https://whatcanirunai.vercel.app/models/gemma-4-31b-it/) | 18.7 GBQ4\_K\_M | **22.5 GB** | — | 32 GB | [Intel Arc Pro B70 32GB](https://whatcanirunai.vercel.app/hardware/arc-pro-b70/)$949 list | [Mac mini M5 Pro 64GB](https://whatcanirunai.vercel.app/hardware/mac-mini-m5-pro-64/)$3,199 list | | deepseek-r1:32b[DeepSeek-R1-Distill-Qwen-32B](https://whatcanirunai.vercel.app/models/deepseek-r1-distill-qwen-32b/) | 19.9 GBQ4\_K\_M | **23.8 GB** | 31.3 GB | 32 GB | [Intel Arc Pro B70 32GB](https://whatcanirunai.vercel.app/hardware/arc-pro-b70/)$949 list | [Mac mini M5 Pro 64GB](https://whatcanirunai.vercel.app/hardware/mac-mini-m5-pro-64/)$3,199 list | | qwen2.5-coder:32b[Qwen2.5-Coder-32B-Instruct](https://whatcanirunai.vercel.app/models/qwen2.5-coder-32b-instruct/) | 19.9 GBQ4\_K\_M | **23.8 GB** | 31.3 GB | 32 GB | [Intel Arc Pro B70 32GB](https://whatcanirunai.vercel.app/hardware/arc-pro-b70/)$949 list | [Mac mini M5 Pro 64GB](https://whatcanirunai.vercel.app/hardware/mac-mini-m5-pro-64/)$3,199 list | | qwen3:32b[Qwen3-32B](https://whatcanirunai.vercel.app/models/qwen3-32b/) | 20.2 GBQ4\_K\_M | **24.2 GB** | 31.7 GB | 32 GB | [Intel Arc Pro B70 32GB](https://whatcanirunai.vercel.app/hardware/arc-pro-b70/)$949 list | [Mac mini M5 Pro 64GB](https://whatcanirunai.vercel.app/hardware/mac-mini-m5-pro-64/)$3,199 list | | llama3.3:70b[Llama-3.3-70B-Instruct](https://whatcanirunai.vercel.app/models/llama-3.3-70b-instruct/) | 42.5 GBQ4\_K\_M | **49.9 GB** | 59.3 GB | 64 GB | [NVIDIA RTX Pro 6000 Blackwell 96GB](https://whatcanirunai.vercel.app/hardware/rtx-pro-6000-blackwell/)$8,565 list | [Framework Desktop (Ryzen AI Max+ 395) 128GB](https://whatcanirunai.vercel.app/hardware/framework-desktop-128/)$3,449 list | | deepseek-r1:70b[DeepSeek-R1-Distill-Llama-70B](https://whatcanirunai.vercel.app/models/deepseek-r1-distill-llama-70b/) | 42.5 GBQ4\_K\_M | **49.9 GB** | 59.3 GB | 64 GB | [NVIDIA RTX Pro 6000 Blackwell 96GB](https://whatcanirunai.vercel.app/hardware/rtx-pro-6000-blackwell/)$8,565 list | [Framework Desktop (Ryzen AI Max+ 395) 128GB](https://whatcanirunai.vercel.app/hardware/framework-desktop-128/)$3,449 list | | qwen2.5:72b[Qwen2.5-72B-Instruct](https://whatcanirunai.vercel.app/models/qwen2.5-72b-instruct/) | 47.4 GBQ4\_K\_M | **55.5 GB** | 64.9 GB | 64 GB | [NVIDIA RTX Pro 6000 Blackwell 96GB](https://whatcanirunai.vercel.app/hardware/rtx-pro-6000-blackwell/)$8,565 list | [Framework Desktop (Ryzen AI Max+ 395) 128GB](https://whatcanirunai.vercel.app/hardware/framework-desktop-128/)$3,449 list | | gpt-oss:120b[gpt-oss-120b](https://whatcanirunai.vercel.app/models/gpt-oss-120b/) | 65.4 GBMXFP4 | **76.2 GB** | 77.2 GB | 96 GB | [NVIDIA RTX Pro 6000 Blackwell 96GB](https://whatcanirunai.vercel.app/hardware/rtx-pro-6000-blackwell/)$8,565 list | [Framework Desktop (Ryzen AI Max+ 395) 128GB](https://whatcanirunai.vercel.app/hardware/framework-desktop-128/)$3,449 list | | qwen3:235b[Qwen3-235B-A22B](https://whatcanirunai.vercel.app/models/qwen3-235b-a22b/) | 142.2 GBQ4\_K\_M | **164.5 GB** | 170.0 GB | 192 GB | None we track | [Mac Studio M5 Ultra 256GB](https://whatcanirunai.vercel.app/hardware/mac-studio-m5-ultra-256/)$9,499 list | | deepseek-r1:671b[DeepSeek-R1-0528](https://whatcanirunai.vercel.app/models/deepseek-r1-0528/) | 404.4 GBQ4\_K\_M | **466.1 GB** | — | 512 GB+ | None we track | None we track | Model file size and quantization from each tag's manifest on registry.ollama.ai, context window from the tag list on ollama.com, retrieved 2026-11-12 (for example [ollama.com/library/qwen3/tags](https://ollama.com/library/qwen3/tags)). "Model file" is the language-model weights; Gemma tags also pull a vision projector (about 0.2–1 GB) that adds memory when you send images. Most default tags are Q4\_K\_M. llama3.2:1b defaults to Q8\_0, mistral-nemo to Q4\_0, and gpt-oss ships as MXFP4\. "—" in the 32k column means we don't have the model's attention config yet; "max Nk ctx" means the tag's context window (from Ollama's tags page) is shorter than 32k. Tags can be updated by Ollama, so sizes may change. ## Which Ollama models fit your GPU ### 8 GB VRAM 14 of these 39 tags fit. Largest: `qwen3:8b` (\~7.0 GB), `deepseek-r1:8b` (\~7.0 GB), `granite3.3:8b` (\~6.7 GB), `llama3.1:8b` (\~6.7 GB), `deepseek-r1:7b` (\~6.4 GB), `qwen2.5-coder:7b` (\~6.4 GB). ### 12 GB VRAM 19 of these 39 tags fit. Largest: `olmo2:13b` (\~10.6 GB), `gemma4:12b` (\~9.5 GB), `gemma3:12b` (\~9.4 GB), `mistral-nemo:12b` (\~9.1 GB), `gemma4:e4b` (\~7.3 GB), `qwen3:8b` (\~7.0 GB). ### 16 GB VRAM 23 of these 39 tags fit. Largest: `qwen3:14b` (\~11.7 GB), `phi4:14b` (\~11.4 GB), `deepseek-r1:14b` (\~11.3 GB), `qwen2.5-coder:14b` (\~11.3 GB), `olmo2:13b` (\~10.6 GB), `gemma4:12b` (\~9.5 GB). ### 24 GB VRAM 27 of these 39 tags fit. Largest: `gemma4:26b` (\~20.6 GB), `gemma3:27b` (\~20.0 GB), `mistral-small3.2:24b` (\~18.5 GB), `gpt-oss:20b` (\~16.9 GB, ✓32k), `qwen3:14b` (\~11.7 GB, ✓32k), `phi4:14b` (\~11.4 GB). Ollama defaults to 32k context on 24–48 GB cards, so ✓32k marks tags that still fit there. ### 32 GB VRAM 33 of these 39 tags fit. Largest: `qwen3:32b` (\~24.2 GB), `deepseek-r1:32b` (\~23.8 GB), `qwen2.5-coder:32b` (\~23.8 GB), `gemma4:31b` (\~22.5 GB), `qwen3:30b` (\~22.3 GB, ✓32k), `qwen3-coder:30b` (\~22.3 GB, ✓32k). Ollama defaults to 32k context on 24–48 GB cards, so ✓32k marks tags that still fit there. ### 48 GB VRAM 33 of these 39 tags fit. Largest: `qwen3:32b` (\~24.2 GB), `deepseek-r1:32b` (\~23.8 GB), `qwen2.5-coder:32b` (\~23.8 GB), `gemma4:31b` (\~22.5 GB), `qwen3:30b` (\~22.3 GB), `qwen3-coder:30b` (\~22.3 GB). Ollama defaults to 256k context on 48 GB+ cards, which can need far more memory than the 4k numbers here; lower it if `ollama ps` shows CPU offload. ### 64 GB VRAM 36 of these 39 tags fit. Largest: `qwen2.5:72b` (\~55.5 GB), `llama3.3:70b` (\~49.9 GB), `deepseek-r1:70b` (\~49.9 GB), `qwen3:32b` (\~24.2 GB), `deepseek-r1:32b` (\~23.8 GB), `qwen2.5-coder:32b` (\~23.8 GB). Ollama defaults to 256k context on 48 GB+ cards, which can need far more memory than the 4k numbers here; lower it if `ollama ps` shows CPU offload. ### 128 GB VRAM 37 of these 39 tags fit. Largest: `gpt-oss:120b` (\~76.2 GB), `qwen2.5:72b` (\~55.5 GB), `llama3.3:70b` (\~49.9 GB), `deepseek-r1:70b` (\~49.9 GB), `qwen3:32b` (\~24.2 GB), `deepseek-r1:32b` (\~23.8 GB). Ollama defaults to 256k context on 48 GB+ cards, which can need far more memory than the 4k numbers here; lower it if `ollama ps` shows CPU offload. ## Ollama default context and why it matters Ollama picks a default context length from your VRAM ([docs](https://github.com/ollama/ollama/blob/main/docs/context-length.mdx)): **4k** under 24 GiB, **32k** on 24–48 GiB, and **256k** at 48 GiB or more. The KV cache grows with context, so a model that fits at 4k can spill to the CPU at 32k. For example, qwen3:30b needs about 22.3 GB at 4k and 25.2 GB at 32k. See [context length and VRAM](https://whatcanirunai.vercel.app/context-length/) for the math. To set it yourself: `OLLAMA_CONTEXT_LENGTH=8192 ollama serve`, or `/set parameter num_ctx 8192` inside `ollama run`. Check the result with `ollama ps`. The PROCESSOR column should read **100% GPU**. A split like "48%/52% CPU/GPU" means part of the model is on the CPU and speed drops a lot. ## Quick answers How much VRAM do I need for Ollama? It depends on the model. Ollama's default tags are mostly 4-bit (Q4\_K\_M), so plan on roughly the model file size plus 15% and about 1 GB at a 4k context: llama3.1:8b \~6.7 GB, gpt-oss:20b \~16.9 GB, qwen3:32b \~24.2 GB, llama3.3:70b \~49.9 GB. 8 GB runs 7B–8B models, 16 GB runs 14B models, 24 GB runs gpt-oss:20b and 24B–27B models, 32 GB runs 30B–32B models, and a 70B tag needs about 49.9 GB, so 64 GB or more. How much VRAM does gpt-oss:20b need in Ollama? About 16.9 GB at a 4k context by our conservative estimate and 17.6 GB at 32k. The 13.8 GB model file is OpenAI's native MXFP4 weights. OpenAI [says](https://openai.com/index/introducing-gpt-oss/) it runs within 16 GB of memory, so a 16 GB GPU can work at short context, but a 24 GB card leaves comfortable room. Can I run Ollama with 8 GB of VRAM? Yes. llama3.1:8b needs about 6.7 GB and qwen3:8b needs about 7.0 GB at a 4k context, which fits an 8 GB card. 12B–14B tags (about 9–12 GB) don't fit fully, so Ollama splits them between GPU and CPU, which is much slower. What happens if a model is bigger than my VRAM in Ollama? Ollama offloads the layers that don't fit to system RAM and runs them on the CPU. It still works, but generation slows down a lot. Run ollama ps and check the PROCESSOR column: you want "100% GPU". If it shows a CPU/GPU split, use a smaller tag or shorter context. What context length does Ollama use by default? It depends on your VRAM, per the [Ollama docs](https://github.com/ollama/ollama/blob/main/docs/context-length.mdx): 4k context under 24 GiB, 32k on 24–48 GiB, and 256k at 48 GiB or more. Longer context needs more memory for the KV cache. Set the OLLAMA\_CONTEXT\_LENGTH environment variable (for example OLLAMA\_CONTEXT\_LENGTH=8192) to change it. How do I reduce Ollama's memory use? Pick a smaller tag, lower the context length with OLLAMA\_CONTEXT\_LENGTH or num\_ctx, and quantize the KV cache with OLLAMA\_KV\_CACHE\_TYPE=q8\_0 (about half the cache memory of the default f16, per the [Ollama FAQ](https://github.com/ollama/ollama/blob/main/docs/faq.mdx)). That needs Flash Attention, which Ollama turns on automatically where supported. Can I run Llama 3.3 70B in Ollama on a 24 GB GPU? Not fully. llama3.3:70b is a 42.5 GB download and needs about 49.9 GB at a 4k context. On a 24 GB card Ollama offloads about half of it to the CPU, so it runs slowly. It needs 64 GB or more of GPU memory to run fully on the GPU, for example two 32 GB cards, a 96 GB card or a 128 GB unified-memory machine (see the table above for the cheapest fit). How much memory does deepseek-r1:671b need in Ollama? About 466 GB at a 4k context: the download alone is 404 GB. That is beyond any single GPU or Mac we track. The deepseek-r1:7b, 8b, 14b, 32b and 70b tags are smaller distilled models, not the full R1. [How much VRAM by model size](https://whatcanirunai.vercel.app/how-much-vram/) · [GGUF VRAM by quant (llama.cpp, LM Studio)](https://whatcanirunai.vercel.app/gguf-vram-requirements/) · [VRAM calculator](https://whatcanirunai.vercel.app/vram-calculator/) · [Q4 vs Q8 vs FP16](https://whatcanirunai.vercel.app/quantization/) · [Context length and VRAM](https://whatcanirunai.vercel.app/context-length/) · [Tokens per second](https://whatcanirunai.vercel.app/tokens-per-second/) · [Best GPU for local LLMs](https://whatcanirunai.vercel.app/best-gpu-for-local-llm/) · [Best LLMs by VRAM size](https://whatcanirunai.vercel.app/best-llm-for/) --- Agent guide: https://whatcanirunai.vercel.app/llms.txt · JSON API: https://whatcanirunai.vercel.app/developers/