Model index
Models, without the guesswork.
Each record states its source and verification date. Weight sizes come from published GGUF files; KV-cache figures are conservative estimates from public model configs. Unknown speed data stays unknown instead of becoming a fake benchmark.
| Model | Architecture | Parameters | Tier | Context | Smallest files | Verified |
|---|---|---|---|---|---|---|
| gpt-oss-20b | MoE | 21B total / 3.6B active | Consumer | 131,072 | 12.11 GB MXFP4 | |
| Qwen3.6 35B-A3B | MoE | 35B total / 3B active | Workstation | 262,144 | 22.61 GB MXFP4 | |
| Qwen3.8 27B | Dense | 27B dense | Workstation | 262,144 | 17.39 GB Q4_K_M | |
| Gemma 4 12B Unified | Dense | 11.95B dense | Consumer | 262,144 | 7.15 GB Q4_0 | |
| Gemma 4 E4B | Dense | 4.5B effective / 8B stored | Consumer | 131,072 | 5.97 GB Q4_K_M | |
| Gemma 4 31B | Dense | 30.7B dense | Workstation | 262,144 | 19.52 GB Q4_K_M | |
| gpt-oss-120b | MoE | 117B total / 5.1B active | Workstation | 131,072 | 63.39 GB MXFP4 | |
| DeepSeek V4.1 Flash | MoE | 552B total / 16B active | Server | 1,048,576 | 232.29 GB UD-IQ1_M | |
| Llama 3.2 3B | Dense | 3.2B dense | Consumer | 131,072 | 2.00 GB Q4_K_M | |
| Qwen3 8B | Dense | 8.2B dense | Consumer | 32,768 | 5.20 GB Q4_K_M | |
| Gemma 3 4B | Dense | 4B dense | Consumer | 131,072 | 3.40 GB Q4_K_M | |
| Phi-4 Mini | Dense | 3.8B dense | Consumer | 131,072 | 2.50 GB Q4_K_M | |
| Mistral 7B | Dense | 7.2B dense | Consumer | 32,768 | 4.40 GB Q4_K_M | |
| DeepSeek R1 8B | Dense | 8B dense | Consumer | 131,072 | 4.90 GB Q4_K_M |
Server-tier models are listed for research but excluded from the compatibility calculator. Decimal GB is used throughout. Browse VRAM-tier guides or read the calculation methodology.