AgentWorldBench (OS subset) — Distinct domain subset of AgentWorldBench
| # | Model | Vendor | Best score | Runs | Last seen |
|---|---|---|---|---|---|
| 1 | Claude Opus 4.6 | Anthropic | 70.2 | 1 | 2026-06-24 |
| 2 | GPT-5.4 | OpenAI | 68.6 | 1 | 2026-06-24 |
| 3 | Qwen AgentWorld 397B A17B | Alibaba | 67.9 | 1 | 2026-06-24 |
| 4 | Gemini 3.1 Pro | 66.9 | 1 | 2026-06-24 | |
| 5 | Claude Opus 4.8 | Anthropic | 66.6 | 1 | 2026-06-24 |
| 6 | Qwen AgentWorld 35B A3B | Alibaba | 65.9 | 1 | 2026-06-24 |
| 7 | DeepSeek-V4-Pro | DeepSeek | 63.7 | 1 | 2026-06-24 |
| 8 | Claude Sonnet 4.6 | Anthropic | 63.2 | 1 | 2026-06-24 |
| 9 | Qwen3.5 397B A17B | Alibaba | 60.9 | 1 | 2026-06-24 |
| 10 | Kimi K2.6 | Moonshot | 60.8 | 1 | 2026-06-24 |
| 11 | Qwen3.6 Plus | Alibaba | 60.3 | 1 | 2026-06-24 |
| 12 | GLM-5.1 | Z.ai | 59.1 | 1 | 2026-06-24 |
| 13 | MiniMax M2.7 | MiniMax | 57.7 | 1 | 2026-06-24 |
| 14 | Qwen3.5 35B A3B | Alibaba | 56.3 | 1 | 2026-06-24 |
Best tracked score per model (default configuration; source-attributed and verification-tiered). Open a model for its full benchmark surface, provenance and pricing.