FireBench - Item Ranking: leaderboard
Metric: Share of the 200 item ranking samples (%) in which the model performs the expected action: returning the top-N rows of a realistic table verbatim, sorted by a designated attribute, verified programmatically; FireBench enterprise and API instruction following; higher is better. Source: arxiv.org. Saturation forecast: Estimated already saturated. 11 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | GPT-5.1 (Medium) | 93 | #131 (GPT-5.1) |
| 2 | DeepSeek V3.1 Terminus | 92.5 | #212 |
| 3 | Qwen 3 235B A22B 2507 (Thinking) | 76 | #253 (Qwen 3 235B A22B 2507) |
| 4 | Kimi K2 (Thinking) | 72.5 | #236 (Kimi K2) |
| 5 | GPT-OSS-120B | 71 | #330 |
| 6 | Claude Sonnet 4.5 | 63 | #138 |
| 7 | Qwen 3 235B A22B 2507 Instruct | 38 | #291 |
| 8 | GPT-4.1 | 32.5 | #240 |
| 9 | Llama 4 Maverick Instruct | 27.9 | #439 |
| 10 | GPT-5.1 Instant | 16 | #277 |
| 11 | Kimi K2 Instruct (0905) | 8.5 | #247 |
No result here: #3 Claude Opus 5.5, #5 GPT-6 Astra, #8 Claude Fable 5.1.
Interactive version: theaggregate.ai/benchmark?slug=firebench-item-ranking · How It Works · Data refreshed daily, snapshot 2026-10-11.