PrismML Bonsai 2 Small - Tau2-Bench: leaderboard
PrismML’s comparison of small Bonsai 2 and Qwen3.5 models on conversational tool use across airline, retail and telecom tasks. Models use thinking with the same evaluation harness and sampling settings; the published scores are percentages.
Metric: Single-trial pass@1 (%). Source: raw.githubusercontent.com. Saturation forecast: Around January 2027. 6 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Qwen 3.5 9B | 85.25 | #387 |
| 2 | Qwen 3.5 4B | 81.65 | #496 |
| 3 | Qwen 3.5 2B | 66.55 | #1146 |
No result here: #3 Claude Opus 5.5, #5 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=prismml-bonsai-2-small-tau2-bench · How It Works · Data refreshed daily, snapshot 2026-10-29.