Nanbeige4.3 Release - Toolathlon-Verified: leaderboard
Toolathlon-Verified tool-use workflow tasks, with no max_tokens cap on individual model requests. Score (%), higher is better. Nanbeige4.3-33B release card (Hugging Face, 2026-10-25): Nanbeige ran all four models (Nanbeige4.3-33B, Qwen3.8-27B, Muse-Glimmer-30B, Gemma4-31B) at their highest thinking effort with one harness and one set of inference settings per benchmark; effort names, sampling and run counts are not stated.
Metric: Score (%). Source: huggingface.co. Saturation forecast: Around February 2027. 4 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Nanbeige4.3-33B | 70.4 | #135 |
| 2 | Qwen 3.8 27B | 68.5 | #123 |
| 3 | Muse Glimmer 30B | 47.2 | #191 |
| 4 | Gemma 4 31B | 27.8 | #212 |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=nanbeige4-3-release-toolathlon-verified · How It Works · Data refreshed daily, snapshot 2026-11-05.