Open weights
Only rows with openWeights=true. Null is not a yes. Quality leads; unrated sits below, never as a win. Sizes this page does not have are not guessed.
The self-host fitter skips models with no sourced size. MoE memory uses total parameters, not active. This list is the catalogue flag, not a Hugging Face scrape.
| Model | Quality | Input $/M | Context | Licence |
|---|---|---|---|---|
| Kimi K3 | 1476.3 | $3.00 | 1.0M | modified-mit |
| GLM 5.2 | 1465.4 | $1.19 | 1.0M | mit |
| Gemma 4 31B | 1441.7 | $0.090 | 262K | gemma |
| Qwen3.8 27B | 1440.8 | $0.425 | 1M | apache-2.0 |
| DeepSeek V4 Pro 0423 | 1439.2 | $0.870 | 1.0M | — |
| MiniMax M3 | 1434.8 | $0.300 | 1.0M | mit |
| DeepSeek V4 Flash 0423 | 1431.6 | $0.081 | 1.0M | — |
| DeepSeek V3.2 | 1424.6 | $0.260 | 164K | mit |
| gpt-oss-120b | 1365.6 | $0.037 | 131K | apache-2.0 |
| Olmo 3 32B Think | 1298.5 | $0.150 | 66K | apache-2.0 |
| Granite 4.1 8B | 1291.6 | $0.050 | 131K | apache-2.0 |
| Phi 4 | 1216.8 | $0.070 | 16K | mit |
| Llama 4 Scout | — | $0.110 | 1.3M | llama-4-community |
| Mistral Small 4 | — | $0.150 | 262K | apache-2.0 |
| Llama 4 Maverick | — | $0.200 | 1.0M | llama-4-community |
| Muse Glimmer 30B | — | $0.350 | 131K | — |
| Nemotron 3 Ultra | — | $0.600 | 512K | nvidia-open-model |
| Kimi K2.7 Code | — | $0.670 | 262K | modified-mit |
| Qwen3.8 2.4T A95B | — | $2.00 | 1.0M | apache-2.0 |
| Command A | — | $2.50 | 256K | cc-by-nc-4.0 |
8 have no independent score. They are listed after the ranked rows.