Free tool · runs in your browser

Which model fits my GPU?

Pick your hardware and what you want to do. We'll list Ollama models that fit, how much memory they need, and a rough speed estimate.

Turn on JavaScript to use the picker. The full model table is below.

How we estimate: memory needed = Ollama's download size for the 4-bit tag + about 1.5 GB for context and overhead. Speed is an estimate from typical memory bandwidth for your hardware class (generation speed is mostly limited by how fast the model's weights can be read). Real speeds vary with your exact card, drivers, context length and settings. Check with ollama ps that the model shows "100% GPU".

All models in the picker

Sizes are Ollama library download sizes for the default 4-bit tag, checked 2026-10-17. Model facts come from each model's Ollama page and model card.

Ollama tagFamilyParametersQuantDownloadEst. memoryContextGood forSources
gemma3:1bGemma 31BQ4_K_M0.8 GB2.3 GB32Kgeneral, lowpowerOllama · Model card
gemma3:4bGemma 34BQ4_K_M3.4 GB4.9 GB128Kgeneral, vision, lowpowerOllama · Model card
gemma3:12bGemma 312BQ4_K_M8.2 GB9.7 GB128Kgeneral, vision, writingOllama · Model card
gemma3:27bGemma 327BQ4_K_M17 GB18.5 GB128Kgeneral, vision, writingOllama · Model card
gemma4:e2bGemma 4E2B (effective 2B)Q4_K_M4.6 GB6.1 GB128Kgeneral, vision, lowpowerOllama · Model card
gemma4:e4bGemma 4E4B (effective 4B)Q4_K_M6.6 GB8.1 GB128Kgeneral, vision, lowpowerOllama · Model card
gemma4:12bGemma 412BQ4_K_M8 GB9.5 GB256Kgeneral, vision, writing, reasoningOllama · Model card
gemma4:26bGemma 426B MoE (~4B active)Q4_K_M18 GB19.5 GB256Kgeneral, vision, reasoning, codingOllama · Model card
gemma4:31bGemma 431BQ4_K_M20 GB21.5 GB256Kgeneral, vision, reasoning, writingOllama · Model card
qwen3:4bQwen34BQ4_K_M2.5 GB4.0 GB256Kgeneral, reasoning, lowpowerOllama · Model card
qwen3:8bQwen38BQ4_K_M5.2 GB6.7 GB40Kgeneral, reasoningOllama · Model card
qwen3:14bQwen314BQ4_K_M9.3 GB10.8 GB40Kgeneral, reasoningOllama · Model card
qwen3:30bQwen330B MoE (3B active)Q4_K_M19 GB20.5 GB256Kgeneral, reasoning, ragOllama · Model card
qwen3.5:4bQwen3.54BQ4_K_M3.3 GB4.8 GB256Kgeneral, vision, lowpowerOllama · Model card
qwen3.5:9bQwen3.59BQ4_K_M6.6 GB8.1 GB256Kgeneral, vision, reasoning, ragOllama · Model card
qwen3.5:27bQwen3.527BQ4_K_M17 GB18.5 GB256Kgeneral, vision, reasoning, coding, ragOllama · Model card
qwen3.5:35bQwen3.535B MoE (3B active)Q4_K_M22 GB23.5 GB256Kgeneral, vision, reasoning, ragOllama · Model card
qwen3.6:27bQwen3.627BQ4_K_M17 GB18.5 GB256Kgeneral, vision, reasoning, coding, ragOllama · Model card
qwen3.6:27b-codingQwen3.627B (coding tune)Q4_K_M18 GB19.5 GB256KcodingOllama · Model card
qwen3.6:35bQwen3.635B MoE (3B active)Q4_K_M24 GB25.5 GB256Kgeneral, vision, reasoning, coding, ragOllama · Model card
qwen3-coder:30bQwen3-Coder30B MoE (3B active)Q4_K_M19 GB20.5 GB256KcodingOllama · Model card
qwen2.5-coder:1.5bQwen2.5-Coder1.5BQ4_K_M1 GB2.5 GB32Kcoding, autocomplete, lowpowerOllama · Model card
qwen2.5-coder:7bQwen2.5-Coder7BQ4_K_M4.7 GB6.2 GB32Kcoding, autocompleteOllama · Model card
qwen2.5-coder:14bQwen2.5-Coder14BQ4_K_M9 GB10.5 GB32KcodingOllama · Model card
qwen3-vl:8bQwen3-VL8BQ4_K_M6.1 GB7.6 GB256KvisionOllama · Model card
llama3.2:1bLlama 3.21BQ4_K_M0.8 GB2.3 GB128Kgeneral, lowpowerOllama · Model card
llama3.2:3bLlama 3.23BQ4_K_M2 GB3.5 GB128Kgeneral, lowpowerOllama · Model card
llama3.1:8bLlama 3.18BQ4_K_M4.9 GB6.4 GB128Kgeneral, ragOllama · Model card
llama3.3:70bLlama 3.370BQ4_K_M43 GB44.5 GB128Kgeneral, writingOllama · Model card
ministral-3:3bMinistral 33BQ4_K_M3 GB4.5 GB256Kgeneral, vision, lowpowerOllama · Model card
ministral-3:8bMinistral 38BQ4_K_M6 GB7.5 GB256Kgeneral, vision, ragOllama · Model card
ministral-3:14bMinistral 314BQ4_K_M9.1 GB10.6 GB256Kgeneral, vision, rag, writingOllama · Model card
mistral-nemoMistral NeMo12BQ4_K_M7.1 GB8.6 GB128Kgeneral, writing, ragOllama · Model card
mistral-small3.2Mistral Small 3.224BQ4_K_M15 GB16.5 GB128Kgeneral, vision, writing, ragOllama · Model card
magistralMagistral Small24BQ4_K_M14 GB15.5 GB39KreasoningOllama · Model card
devstral-small-2Devstral Small 224BQ4_K_M15 GB16.5 GB384KcodingOllama · Model card
deepseek-r1:1.5bDeepSeek-R1 (distill)1.5BQ4_K_M1.1 GB2.6 GB128Kreasoning, lowpowerOllama · Model card
deepseek-r1:8bDeepSeek-R1 (0528 Qwen3 distill)8BQ4_K_M5.2 GB6.7 GB128KreasoningOllama · Model card
deepseek-r1:14bDeepSeek-R1 (distill)14BQ4_K_M9 GB10.5 GB128KreasoningOllama · Model card
deepseek-r1:32bDeepSeek-R1 (distill)32BQ4_K_M20 GB21.5 GB128KreasoningOllama · Model card
phi4-miniPhi-4-mini3.8BQ4_K_M2.5 GB4.0 GB128Kgeneral, reasoning, lowpowerOllama · Model card
phi4Phi-414BQ4_K_M9.1 GB10.6 GB16KreasoningOllama · Model card
phi4-reasoningPhi-4-reasoning14BQ4_K_M11 GB12.5 GB32KreasoningOllama · Model card
gpt-oss:20bgpt-oss21B MoE (3.6B active)MXFP414 GB15.5 GB128Kgeneral, reasoning, codingOllama · Model card
gpt-oss:120bgpt-oss117B MoE (5.1B active)MXFP465 GB66.5 GB128Kgeneral, reasoning, codingOllama · Model card
granite4.2:3bGranite 4.23BQ4_K_M2.2 GB3.7 GB128Kgeneral, rag, lowpowerOllama · Model card
granite4.2:8bGranite 4.28BQ4_K_M5.3 GB6.8 GB128Kgeneral, ragOllama · Model card
nemotron-3-nano:4bNemotron 3 Nano4BQ4_K_M2.8 GB4.3 GB256Kgeneral, reasoning, lowpowerOllama · Model card
nomic-embed-textNomic Embed137MQ4_K_M0.27 GB0.6 GB2KembedOllama · Model card
qwen3-embedding:0.6bQwen3-Embedding0.6BQ4_K_M0.64 GB0.9 GB32KembedOllama · Model card

Get the Local AI weekly

New models worth running, what fits which GPU, and setup fixes. One short email a week, free. Unsubscribe any time.

Lila Forge Kits

Local-LLM App Starter

A tested starter for building your own app on Ollama: chat, RAG and a GPU-fit checker. Skip the setup hours.

kits.lilaforge.com → VibeEngine

Best local LLM tools, ranked

Ollama, LM Studio, Open WebUI and llama.cpp scored from public evidence, with top picks.

vibeengine.lilaforge.com →