IFEval — Explicit averaged-metric variant, IBM Granite cards; distinct axis from single strict-prompt metric.
| # | Model | Vendor | Best score | Runs | Last seen |
|---|---|---|---|---|---|
| 1 | Claude 3.5 Sonnet | Anthropic | 90.1 | 1 | 2026-06-06 |
| 2 | Granite 4.1 30B | IBM | 89.7 | 1 | 2026-06-15 |
| 3 | Gemini 1.5 Pro | 89.4 | 1 | 2026-06-06 | |
| 4 | MiniMax Text 01 | MiniMax | 89.1 | 1 | 2026-06-06 |
| 5 | Gemini 2.0 Flash Exp | 88.4 | 1 | 2026-06-06 | |
| 6 | DeepSeek-V3 | DeepSeek | 87.3 | 1 | 2026-06-06 |
| 7 | Qwen2.5 Instruct 72B | Alibaba | 87.2 | 1 | 2026-06-06 |
| 8 | Granite 4.1 8B | IBM | 87.1 | 1 | 2026-06-15 |
| 9 | Llama 3.1 Instruct 405B | Meta | 86.4 | 1 | 2026-06-06 |
| 10 | GPT-4o | OpenAI | 84.1 | 1 | 2026-06-06 |
Best tracked score per model (default configuration; source-attributed and verification-tiered). Open a model for its full benchmark surface, provenance and pricing.