IFEval - release-reported, metric unspecified
Chat & Writing
IFEval verifiable instruction-following results from a model release that does not say which of the four standard rates (prompt- or instruction-level, strict or loose), or which average of them, is reported. Shares the IFEval group without claiming the canonical definition's single 541-prompt proportion.
Top models (higher is better)
| Model | Score |
|---|---|
| LFM2.5-24B-A2B | 96.8 |
| LFM3-1.7B | 96.7 |
| LFM2.5-2.6B | 95.4 |
| Nemotron 3.5 Lightning 30B A3B (precision unspecified) | 94.2 |
| Qwen3.6 35B A3B | 93.1 |
| Qwen3.5-4B | 92.5 |
| Qwen3.5-9B | 92.1 |
| Gemma 4 26B A4B IT | 92.0 |
| LFM2.5-VL-3B | 82.3 |
| LFM2.5-350M | 77.0 |
| LFM2-1.2B | 74.9 |
| Qwen3 1.7B | 74.0 |
| LFM2-700M | 72.2 |
| LFM2.5-230M | 71.7 |
| Qwen3 0.6B | 64.2 |
| Granite 4.0 H 350M | 61.3 |
| LFM2.5-VL-450M | 61.2 |
| Qwen3.5-0.8B | 59.9 |
| Granite 4.0 350M | 53.5 |
| Llama 3.2 1B Instruct | 52.4 |
| LFM2-VL-450M | 51.8 |