LOCA-bench (96K Tokens): leaderboard
Metric: Task success rate (%) over the 75 instances whose environment description is 96K tokens long; 15 seed agentic tasks adapted from Toolathlon with five seeded environment states each, on local mock servers (Canvas, email, calendar, BigQuery, Snowflake, WooCommerce, spreadsheets and more); ReAct scaffold, each model at its maximum context window with the oldest context truncated beyond it; success checked by rule-based scripts on the final environment state; higher is better. Source: arxiv.org. Saturation forecast: Around January 2027. 7 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Claude Opus 4.5 | 45.3 | #79 |
| 2 | GPT-5.2 (Medium) | 44 | #105 (GPT-5.2) |
| 3 | Gemini 3 Flash | 32 | #93 |
| 4 | MiniMax-M2.1 | 22.7 | #283 |
| 5 | DeepSeek V3.2 (Thinking) | 16 | #198 (DeepSeek V3.2) |
| 6 | GLM-4.7 | 14.7 | #185 |
| 7 | Kimi K2 (Thinking) | 13.3 | #236 (Kimi K2) |
No result here: #3 Claude Opus 5.5, #5 GPT-6 Astra, #8 Claude Fable 5.1.
Interactive version: theaggregate.ai/benchmark?slug=loca-bench-96k-tokens · How It Works · Data refreshed daily, snapshot 2026-10-11.