IMO-AnswerBench leaderboard
As of November 6, 2026, among the results listed here, Z.ai GLM 5.2 has the highest reported IMO-AnswerBench score, 91, published by Z.ai, ahead of MoonshotAI Kimi K2.6 at 86. 13 models have a IMO-AnswerBench result in our records.
Scores as the model makers published them, not re-run by us; each row links to its announcement.
13 models ranked · 5 makers
| Rank | Model | Score | Setting | Reported by | Date |
|---|---|---|---|---|---|
| 1 | GLM 5.2 | 91 | Not published | Z.ai | June 16, 2026 (model release) |
| 2 | Kimi K2.6 | 86 | Not published | Moonshot AI | April 20, 2026 (model release) |
| 2 | Qwen3.7 Plus | 86 | Not published | Qwen | May 21, 2026 |
| 4 | Qwen3.6 Plus | 83.8 | Not published | Qwen | May 21, 2026 |
| 4 | GLM 5.1 | 83.8 | Not published | Z.ai | April 7, 2026 (model release) |
| 6 | GLM 5 | 82.5 | Not published | Z.ai | February 11, 2026 (model release) |
| 7 | GLM 4.7 | 82 | Not published | Z.ai | December 22, 2025 (model release) |
| 8 | LongCat 2.0 | 81.8 | Not published | Meituan | July 20, 2026 (model release) |
| 8 | Kimi K2.5 | 81.8 | Thinking | Moonshot AI | January 27, 2026 (model release) |
| 10 | Kimi K2 Thinking | 78.6 | Not published | Moonshot AI | November 6, 2025 (model release) |
| 11 | GLM 4.6V | 73.5 | Not published | Z.ai | December 22, 2025 (model release) |
| 12 | Ling 3.0 Tiny | 71 | Thinking | inclusionAI | August 6, 2026 (model release) · read from the maker's chart |
| 13 | Kimi K2 0905 | 45.8 | Not published | Moonshot AI | November 6, 2025 (model release) |
Other reported runs
| Model | Score | Setting | Reported by | Date |
|---|---|---|---|---|
| Kimi K2.5 | 81.8 | Not published | Moonshot AI | April 20, 2026 (model release) |
How much to trust these numbers
- Source
- Reported by the model makers themselves. Not re-run by LLMBenchmark or anyone independent.
- Settings
- Settings differ between makers. Check the Setting column before comparing two scores.
- Spread
- 91 to 45.8 across 13 models.
- Freshness
- Newest result published August 6, 2026; oldest November 6, 2025.
- Charts
- Some values were read from charts in the announcements. Two independent readings had to match exactly.
What IMO-AnswerBench measures
IMO-AnswerBench asks models to solve 400 Olympiad mathematics problems with verifiable short answers; its score reflects how often their answers are correct. Source: IMO-Bench: Towards Robust Mathematical Reasoning