Which model for RTX 4060 · 8 Go?
8 GB usable for the model. 52 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.
Unlocked by this setup
These models fit on no smaller setup in the catalogue (the next one down offers 6 usable GB), even in their lightest quantization.
| Model | Elo | Quantization | RAM required | Margin | Memory headroom | Run |
|---|---|---|---|---|---|---|
| Granite 4.1 8BIBM · 8B | 1304 | Q4_K_M | 7 Go | +1 Go | tight | Hugging Face ↗ |
| Ministral 8BMistral AI · 8.02B | 1238 | Q4_K_M | 7 Go | +1 Go | tight | Hugging Face ↗ |
| Llama 3.1 8BMeta · 8.03B | 1211 | Q4_K_M | 7 Go | +1 Go | tight | ollama run llama3.1:8b |
| Mistral 7BMistral AI · 7.25B | 1110 | Q4_K_M | 7 Go | +1 Go | tight | ollama run mistral:7b |
| Apertus 8BSwiss AI · 8B | — | Q4_K_M | 7 Go | +1 Go | tight | Hugging Face ↗ |
| CodeGemma 7BGoogle · 8.5B | — | Q4_K_M | 7 Go | +1 Go | tight | ollama run codegemma |
| Command R7BCohere · 8B | — | Q4_K_M | 7 Go | +1 Go | tight | ollama run command-r7b |
| DeepSeek-R1 Distill 8BDeepSeek · 8.03B | — | Q4_K_M | 7 Go | +1 Go | tight | ollama run deepseek-r1:8b |
| Granite 3.3 8BIBM · 8.1B | — | Q4_K_M | 7 Go | +1 Go | tight | ollama run granite3.3:8b |
| Hermes 3 8BNous Research · 8.03B | — | Q4_K_M | 7 Go | +1 Go | tight | ollama run hermes3:8b |
| HunyuanVideo 1.5Tencent · 8.3B | — | Q4_K_M | 7 Go | +1 Go | tight | Hugging Face ↗ |
| InternLM3 8BShanghai AI Lab · 8.8B | — | Q4_K_M | 7 Go | +1 Go | tight | Hugging Face ↗ |
| Ministral 3 8BMistral AI · 8B | — | Q4_K_M | 7 Go | +1 Go | tight | Hugging Face ↗ |
| Qwen3 8BAlibaba · 8.2B | — | Q4_K_M | 7 Go | +1 Go | tight | ollama run qwen3:8b |
| EuroLLM 9BUTTER Project · 9.15B | — | Q4_K_M | 8 Go | +0 Go | tight | Hugging Face ↗ |
| Falcon 3 10BTII · 10.3B | — | Q4_K_M | 8 Go | +0 Go | tight | ollama run falcon3:10b |
| GLM-ImageZhipu AI · 9B | — | Q4_K_M | 8 Go | +0 Go | tight | Hugging Face ↗ |
| Mochi 1Genmo · 10B | — | Q4_K_M | 8 Go | +0 Go | tight | Hugging Face ↗ |
Also fit on a smaller setup
| Model | Elo | Quantization | RAM required | Margin | Memory headroom | Run |
|---|---|---|---|---|---|---|
| Gemma 3n E4BGoogle · 4B | 1318 | Q4_K_MQ8_0 also fits · 7 Go | 5 Go | +3 Go | comfortable | ollama run gemma3n:e4b |
| Gemma 3 4BGoogle · 4B | 1303 | Q4_K_MQ8_0 also fits · 7 Go | 5 Go | +3 Go | comfortable | ollama run gemma3:4b |
| Llama 3.2 3BMeta · 3.21B | 1167 | Q4_K_MQ8_0 also fits · 6 Go | 4 Go | +4 Go | comfortable | ollama run llama3.2:3b |
| Llama 3.2 1BMeta · 1.24B | 1111 | Q4_K_MFP16 also fits · 5 Go | 3 Go | +5 Go | comfortable | ollama run llama3.2:1b |
Not yet ranked by the Arena
No Arena Elo linked: recent release, model absent from the Arena, or a match we have not made yet.
| Model | Elo | Quantization | RAM required | Margin | Memory headroom | Run |
|---|---|---|---|---|---|---|
| Qwen3 0.6BAlibaba · 0.6B | — | Q4_K_MFP16 also fits · 3 Go | 2 Go | +6 Go | comfortable | ollama run qwen3:0.6b |
| Gemma 3 1BGoogle · 1B | — | Q4_K_MFP16 also fits · 4 Go | 3 Go | +5 Go | comfortable | ollama run gemma3:1b |
| Phi-4 MiniMicrosoft · 3.8B | — | Q4_K_MQ8_0 also fits · 7 Go | 4 Go | +4 Go | comfortable | ollama run phi4-mini |
| Qwen3 4BAlibaba · 4.02B | — | Q4_K_MQ8_0 also fits · 7 Go | 5 Go | +3 Go | comfortable | ollama run qwen3:4b |
| Gemma 4 E2BGoogle · 2B | — | FP16 | 6 Go | +2 Go | tight | Hugging Face ↗ |
| Granite 4.1 3BIBM · 3B | — | Q8_0 | 6 Go | +2 Go | tight | Hugging Face ↗ |
| Lucie 7BOpenLLM France · 6.7B | — | Q4_K_M | 6 Go | +2 Go | tight | Hugging Face ↗ |
| Ministral 3 3BMistral AI · 3B | — | Q8_0 | 6 Go | +2 Go | tight | Hugging Face ↗ |
| OLMo 3 7BAllen AI · 7B | — | Q4_K_M | 6 Go | +2 Go | tight | Hugging Face ↗ |
| SmolLM3 3BHugging Face · 3.1B | — | Q8_0 | 6 Go | +2 Go | tight | Hugging Face ↗ |
| Gemma 4 E4BGoogle · 4B | — | Q8_0 | 7 Go | +1 Go | tight | Hugging Face ↗ |
| MiniCPM5-2BOpenBMB · 2.5B | — | FP16 | 7 Go | +1 Go | tight | Hugging Face ↗ |
Image, video, audio and 3D
These models do not generate text, so the text Arena does not rank them. Indicative order, not a quality ranking.
| Model | Elo | Quantization | RAM required | Margin | Memory headroom | Run |
|---|---|---|---|---|---|---|
| Kokoro-82MHexgrad · 0.08B | — | FP16 | 2 Go | +6 Go | comfortable | Hugging Face ↗ |
| ChatterboxResemble AI · 0.5B | — | FP16 | 3 Go | +5 Go | comfortable | Hugging Face ↗ |
| Parakeet TDT 0.6B v3NVIDIA · 0.6B | — | FP16 | 3 Go | +5 Go | comfortable | Hugging Face ↗ |
| TripoSRStability AI · 0.5B | — | FP16 | 3 Go | +5 Go | comfortable | Hugging Face ↗ |
| Stable Fast 3DStability AI · 1B | — | FP16 | 4 Go | +4 Go | comfortable | Hugging Face ↗ |
| TRELLISMicrosoft · 1.2B | — | FP16 | 5 Go | +3 Go | comfortable | Hugging Face ↗ |
| Whisper large-v3OpenAI · 1.55B | — | FP16 | 5 Go | +3 Go | comfortable | Hugging Face ↗ |
| Dia 1.6BNari Labs · 1.6B | — | FP16 | 6 Go | +2 Go | tight | Hugging Face ↗ |
| Hunyuan3D 2.1Tencent · 3B | — | Q8_0 | 6 Go | +2 Go | tight | Hugging Face ↗ |
| Ming-Image 0.1 DesignAnt Group · 6B | — | Q4_K_M | 6 Go | +2 Go | tight | Hugging Face ↗ |
| Orpheus 3BCanopy Labs · 3B | — | Q8_0 | 6 Go | +2 Go | tight | Hugging Face ↗ |
| Voxtral Mini 3BMistral AI · 3B | — | Q8_0 | 6 Go | +2 Go | tight | Hugging Face ↗ |
| YuE2-3Bm-a-p (Multimodal Art Projection) · 3B | — | Q8_0 | 6 Go | +2 Go | tight | Hugging Face ↗ |
| Z-Image TurboAlibaba · 6B | — | Q4_K_M | 6 Go | +2 Go | tight | Hugging Face ↗ |
| FLUX.2 [klein] 4BBlack Forest Labs · 4B | — | Q8_0 | 7 Go | +1 Go | tight | Hugging Face ↗ |
| TRELLIS.2Microsoft · 4B | — | Q8_0 | 7 Go | +1 Go | tight | Hugging Face ↗ |
| CogVideoX 5BZhipu AI · 5B | — | Q8_0 | 8 Go | +0 Go | tight | Hugging Face ↗ |
| Wan 2.2 5BAlibaba · 5B | — | Q8_0 | 8 Go | +0 Go | tight | Hugging Face ↗ |
An “ollama run” command without a suffix pulls the Q4_K_M build from the Ollama library: the row then shows that quantization and its memory, and flags whether FP16 or Q8_0 would also fit. Exception: a model published in a native format (MXFP4 for gpt-oss, FP8 for DeepSeek V4.1) exists in one build only, which is what the command loads. via Hugging Face or an explicit Ollama tag.
Share this verdict
One link, the same shortlist, the same reasons.