With 32 GB VRAM, assuming 16 GB system RAM for partial offload, these are the best open models that fit. The top pick is Qwen3 32B at Q4_K_M (~22.2 GB, ~54 tok/s): ollama run qwen3:32b.
Best for coding: Qwen3-Coder 30B-A3B (MoE) (qwen3-coder:30b), Qwen3.8 27B (qwen3.8:27b), Mistral Small 24B (mistral-small:24b-instruct-2501-q8_0).
Runtimes available for this hardware: Ollama, LM Studio, llama.cpp, vLLM.
Open the interactive checker for RTX 5090 32GB| # | Model | Quant | Memory needed | Fit | Speed | Command |
|---|---|---|---|---|---|---|
| 1 | Qwen3 32B | Q4_K_M | 22.2 GB | fits in fast memory | ~54 tok/s | ollama run qwen3:32b |
| 2 | DeepSeek R1 Distill Qwen 32B | Q4_K_M | 22.2 GB | fits in fast memory | ~54 tok/s | ollama run deepseek-r1:32b |
| 3 | Gemma 3 27B | Q4_K_M | 19 GB | fits in fast memory | ~63 tok/s | ollama run gemma3:27b |
| 4 | Clef 27B (Cloudflare, decision) | Default | 19.9 GB | fits in fast memory | ~60 tok/s | ollama run clef:27b |
| 5 | Qwen3.8 27B | Default | 19.9 GB | fits in fast memory | ~60 tok/s | ollama run qwen3.8:27b |
| 6 | Mistral Small 24B | Q8_0 | 26.8 GB | fits in fast memory | ~43 tok/s | ollama run mistral-small:24b-instruct-2501-q8_0 |
| 7 | Qwen3 30B-A3B (MoE) | Q4_K_M | 20.4 GB | fits in fast memory | ~262 tok/s | ollama run qwen3:30b |
| 8 | Qwen3-Coder 30B-A3B (MoE) | Q4_K_M | 20.4 GB | fits in fast memory | ~262 tok/s | ollama run qwen3-coder:30b |
| 9 | gpt-oss 20B (MoE, MXFP4) | MXFP4 | 15.1 GB | fits in fast memory | ~223 tok/s | ollama run gpt-oss:20b |
| 10 | Mixtral 8x7B (MoE) | Q4_K_M | 29.8 GB | fits in fast memory | ~70 tok/s | ollama run mixtral:8x7b-instruct-v0.1-q4_K_M |
| 11 | Qwen3 14B | Q8_0 | 17.3 GB | fits in fast memory | ~67 tok/s | ollama run qwen3:14b-q8_0 |
| 12 | DeepSeek R1 Distill Qwen 14B | Q8_0 | 17.3 GB | fits in fast memory | ~67 tok/s | ollama run deepseek-r1:14b-qwen-distill-q8_0 |
Estimates, not benchmarks. Model sizes are ollama.com download sizes; memory needed adds headroom for ~8k tokens of context; speeds are rough figures from memory bandwidth. Data updated 2026-10-14.
A project by OnDevice Labs · llms.txt