Which model for RTX 5090 · 32 Go?
32 GB usable for the model. 83 open-weight models fit in memory.
The margin goes to context: count ~1 GB per 8k tokens for a 7B model. A small margin means a short context.
| Model | Elo | Quantization | RAM required | Margin | Speed | Run |
|---|---|---|---|---|---|---|
| Gemma 4 31BGoogle · 31B | 1451 | Q4_K_M | 21 Go | +11 Go | comfortable | Hugging Face ↗ |
| Gemma 4 26B-A4BGoogle · 26B | 1438 | Q4_K_M | 18 Go | +14 Go | comfortable | Hugging Face ↗ |
| Gemma 3 27BGoogle · 27B | 1365 | Q4_K_M | 19 Go | +13 Go | comfortable | ollama run gemma3:27b |
| Qwen3 32BAlibaba · 32.8B | 1347 | Q4_K_M | 23 Go | +9 Go | tight | ollama run qwen3:32b |
| Gemma 3 12BGoogle · 12B | 1342 | FP16 | 28 Go | +4 Go | tight | ollama run gemma3:12b |
| QwQ 32BAlibaba · 32.8B | 1336 | Q4_K_M | 23 Go | +9 Go | tight | ollama run qwq |
| Gemma 3n E4BGoogle · 8B | 1317 | FP16 | 20 Go | +12 Go | comfortable | ollama run gemma3n:e4b |
| gpt-oss-20bOpenAI · 21B | 1317 | Q8_0 | 27 Go | +5 Go | tight | ollama run gpt-oss:20b |
| Nemotron 3 Nano 30B-A3BNVIDIA · 30B | 1314 | Q4_K_M | 21 Go | +11 Go | comfortable | Hugging Face ↗ |
| Granite 4.1 8BIBM · 8B | 1306 | FP16 | 20 Go | +12 Go | comfortable | Hugging Face ↗ |
| Gemma 3 4BGoogle · 4B | 1303 | FP16 | 11 Go | +21 Go | comfortable | ollama run gemma3:4b |
| Qwen2.5-Coder 32BAlibaba · 32.8B | 1270 | Q4_K_M | 23 Go | +9 Go | tight | ollama run qwen2.5-coder:32b |
| Phi-4Microsoft · 14.7B | 1256 | Q8_0 | 19 Go | +13 Go | comfortable | ollama run phi4 |
| Phi-4 ReasoningMicrosoft · 14.7B | 1256 | Q8_0 | 19 Go | +13 Go | comfortable | ollama run phi4-reasoning |
| OLMo 2 32BAllen AI · 32B | 1251 | Q4_K_M | 22 Go | +10 Go | comfortable | Hugging Face ↗ |
| CodestralMistral AI · 22.2B | 1240 | Q8_0 | 28 Go | +4 Go | tight | ollama run codestral |
| Ministral 8BMistral AI · 8.02B | 1237 | FP16 | 20 Go | +12 Go | comfortable | Hugging Face ↗ |
| Llama 3.1 8BMeta · 8.03B | 1211 | FP16 | 20 Go | +12 Go | comfortable | ollama run llama3.1:8b |
| Mixtral 8x7BMistral AI · 46.7B | 1197 | Q4_K_M | 31 Go | +1 Go | tight | ollama run mixtral:8x7b |
| Llama 3.2 3BMeta · 3.21B | 1166 | FP16 | 9 Go | +23 Go | comfortable | ollama run llama3.2:3b |
| Llama 3.2 1BMeta · 1.24B | 1111 | FP16 | 5 Go | +27 Go | comfortable | ollama run llama3.2:1b |
| Mistral 7BMistral AI · 7.25B | 1110 | FP16 | 18 Go | +14 Go | comfortable | ollama run mistral:7b |
| Kokoro-82MHexgrad · 0.08B | — | FP16 | 2 Go | +30 Go | comfortable | Hugging Face ↗ |
| ChatterboxResemble AI · 0.5B | — | FP16 | 3 Go | +29 Go | comfortable | Hugging Face ↗ |
| Parakeet TDT 0.6B v3NVIDIA · 0.6B | — | FP16 | 3 Go | +29 Go | comfortable | Hugging Face ↗ |
| Qwen3 0.6BAlibaba · 0.6B | — | FP16 | 3 Go | +29 Go | comfortable | ollama run qwen3:0.6b |
| TripoSRStability AI · 0.5B | — | FP16 | 3 Go | +29 Go | comfortable | Hugging Face ↗ |
| Gemma 3 1BGoogle · 1B | — | FP16 | 4 Go | +28 Go | comfortable | ollama run gemma3:1b |
| Stable Fast 3DStability AI · 1B | — | FP16 | 4 Go | +28 Go | comfortable | Hugging Face ↗ |
| TRELLISMicrosoft · 1.2B | — | FP16 | 5 Go | +27 Go | comfortable | Hugging Face ↗ |
| Whisper large-v3OpenAI · 1.55B | — | FP16 | 5 Go | +27 Go | comfortable | Hugging Face ↗ |
| Dia 1.6BNari Labs · 1.6B | — | FP16 | 6 Go | +26 Go | comfortable | Hugging Face ↗ |
| Gemma 4 E2BGoogle · 2B | — | FP16 | 6 Go | +26 Go | comfortable | Hugging Face ↗ |
| Granite 4.1 3BIBM · 3B | — | FP16 | 9 Go | +23 Go | comfortable | Hugging Face ↗ |
| Hunyuan3D 2.1Tencent · 3B | — | FP16 | 9 Go | +23 Go | comfortable | Hugging Face ↗ |
| Ministral 3 3BMistral AI · 3B | — | FP16 | 9 Go | +23 Go | comfortable | Hugging Face ↗ |
| Orpheus 3BCanopy Labs · 3B | — | FP16 | 9 Go | +23 Go | comfortable | Hugging Face ↗ |
| SmolLM3 3BHugging Face · 3.1B | — | FP16 | 9 Go | +23 Go | comfortable | Hugging Face ↗ |
| Voxtral Mini 3BMistral AI · 3B | — | FP16 | 9 Go | +23 Go | comfortable | Hugging Face ↗ |
| Phi-4 MiniMicrosoft · 3.8B | — | FP16 | 10 Go | +22 Go | comfortable | ollama run phi4-mini |
| FLUX.2 [klein] 4BBlack Forest Labs · 4B | — | FP16 | 11 Go | +21 Go | comfortable | Hugging Face ↗ |
| Gemma 4 E4BGoogle · 4B | — | FP16 | 11 Go | +21 Go | comfortable | Hugging Face ↗ |
| Qwen3 4BAlibaba · 4.02B | — | FP16 | 11 Go | +21 Go | comfortable | ollama run qwen3:4b |
| TRELLIS.2Microsoft · 4B | — | FP16 | 11 Go | +21 Go | comfortable | Hugging Face ↗ |
| CogVideoX 5BZhipu AI · 5B | — | FP16 | 13 Go | +19 Go | comfortable | Hugging Face ↗ |
| Wan 2.2 5BAlibaba · 5B | — | FP16 | 13 Go | +19 Go | comfortable | Hugging Face ↗ |
| Z-Image TurboAlibaba · 6B | — | FP16 | 15 Go | +17 Go | comfortable | Hugging Face ↗ |
| Lucie 7BOpenLLM France · 6.7B | — | FP16 | 17 Go | +15 Go | comfortable | Hugging Face ↗ |
| OLMo 3 7BAllen AI · 7B | — | FP16 | 17 Go | +15 Go | comfortable | Hugging Face ↗ |
| Ministral 3 14BMistral AI · 14B | — | Q8_0 | 18 Go | +14 Go | comfortable | Hugging Face ↗ |
| Qwen3 14BAlibaba · 14.8B | — | Q8_0 | 19 Go | +13 Go | comfortable | ollama run qwen3:14b |
| Qwen3.6 27BAlibaba · 27B | — | Q4_K_M | 19 Go | +13 Go | comfortable | Hugging Face ↗ |
| Wan 2.2 A14BAlibaba · 27B | — | Q4_K_M | 19 Go | +13 Go | comfortable | Hugging Face ↗ |
| Apertus 8BSwiss AI · 8B | — | FP16 | 20 Go | +12 Go | comfortable | Hugging Face ↗ |
| Command R7BCohere · 8B | — | FP16 | 20 Go | +12 Go | comfortable | ollama run command-r7b |
| DeepSeek-Coder-V2 LiteDeepSeek · 15.7B | — | Q8_0 | 20 Go | +12 Go | comfortable | ollama run deepseek-coder-v2:16b |
| DeepSeek-R1 Distill 8BDeepSeek · 8.03B | — | FP16 | 20 Go | +12 Go | comfortable | ollama run deepseek-r1:8b |
| Granite 3.3 8BIBM · 8.1B | — | FP16 | 20 Go | +12 Go | comfortable | ollama run granite3.3:8b |
| Hermes 3 8BNous Research · 8.03B | — | FP16 | 20 Go | +12 Go | comfortable | ollama run hermes3:8b |
| HunyuanVideo 1.5Tencent · 8.3B | — | FP16 | 20 Go | +12 Go | comfortable | Hugging Face ↗ |
| Ministral 3 8BMistral AI · 8B | — | FP16 | 20 Go | +12 Go | comfortable | Hugging Face ↗ |
| Qwen3 8BAlibaba · 8.2B | — | FP16 | 20 Go | +12 Go | comfortable | ollama run qwen3:8b |
| CodeGemma 7BGoogle · 8.5B | — | FP16 | 21 Go | +11 Go | comfortable | ollama run codegemma |
| InternLM3 8BShanghai AI Lab · 8.8B | — | FP16 | 21 Go | +11 Go | comfortable | Hugging Face ↗ |
| StarCoder 2BigCode · 16B | — | Q8_0 | 21 Go | +11 Go | comfortable | ollama run starcoder2:15b |
| Aya Expanse 32BCohere · 32.3B | — | Q4_K_M | 22 Go | +10 Go | comfortable | ollama run aya-expanse:32b |
| EuroLLM 9BUTTER Project · 9.15B | — | FP16 | 22 Go | +10 Go | comfortable | Hugging Face ↗ |
| FLUX.2 [dev]Black Forest Labs · 32B | — | Q4_K_M | 22 Go | +10 Go | comfortable | Hugging Face ↗ |
| GLM-ImageZhipu AI · 9B | — | FP16 | 22 Go | +10 Go | comfortable | Hugging Face ↗ |
| HiDream-I1HiDream · 17B | — | Q8_0 | 22 Go | +10 Go | comfortable | Hugging Face ↗ |
| DeepSeek-R1 Distill 32BDeepSeek · 32.8B | — | Q4_K_M | 23 Go | +9 Go | tight | ollama run deepseek-r1:32b |
| Mochi 1Genmo · 10B | — | FP16 | 24 Go | +8 Go | tight | Hugging Face ↗ |
| Qwen3.6 35B-A3BAlibaba · 35B | — | Q4_K_M | 24 Go | +8 Go | tight | Hugging Face ↗ |
| Falcon 3 10BTII · 10.3B | — | FP16 | 25 Go | +7 Go | tight | ollama run falcon3:10b |
| Qwen-Image 2512Alibaba · 20B | — | Q8_0 | 26 Go | +6 Go | tight | Hugging Face ↗ |
| FLUX.1 [dev]Black Forest Labs · 12B | — | FP16 | 28 Go | +4 Go | tight | Hugging Face ↗ |
| Gemma 4 12BGoogle · 12B | — | FP16 | 28 Go | +4 Go | tight | Hugging Face ↗ |
| LTX-2.3Lightricks · 22B | — | Q8_0 | 28 Go | +4 Go | tight | Hugging Face ↗ |
| Mistral Nemo 12BMistral AI · 12.2B | — | FP16 | 29 Go | +3 Go | tight | ollama run mistral-nemo |
| Devstral SmallMistral AI · 24B | — | Q8_0 | 30 Go | +2 Go | tight | ollama run devstral |
| Magistral SmallMistral AI · 24B | — | Q8_0 | 30 Go | +2 Go | tight | ollama run magistral |
| Mistral Small 3.2Mistral AI · 24B | — | Q8_0 | 30 Go | +2 Go | tight | ollama run mistral-small3.2 |
| Voxtral Small 24BMistral AI · 24B | — | Q8_0 | 30 Go | +2 Go | tight | Hugging Face ↗ |