LongCat 2.5 Release - JobBench: leaderboard
LongCat-2.5 final release detailed native result table, November 2026. External:true cells are explicitly cited other-evaluator results and excluded from this component; self-run numeric cells retained. No Preview/final or release-generation aliases. Detailed table used instead of inconsistent rounded headline charts. JobBench (Main 65): Claude Code harness (v 2.1.280); scoring follows the official JobBench LLM-as-judge protocol (chained rubrics over the submitted deliverables) with gemini-3.6-flash as the judge; temperature = 1.0, top_p = 0.95, thinking enabled. Metric: Score (%).
Metric: Score (%). Source: longcat.ai. Saturation forecast: Rough model projection: around 2026. 3 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | LongCat-2.5 (Thinking) | 65.4 | |
| 2 | DeepSeek V4.1 Flash (Thinking) | 63.2 | #37 (DeepSeek V4.1 Flash) |
| 3 | LongCat-2.0 (Thinking) | 44.4 |
Interactive version: theaggregate.ai/benchmark?slug=longcat-2-5-release-jobbench · How It Works · Data refreshed daily, snapshot 2026-11-05.