LongBench v2 leaderboard
As of November 14, 2026, among the results listed here, Thinking Machines Inkling 2 Small has the highest reported LongBench v2 score, 70.6%, published by Thinking Machines, ahead of Reflection Beam 1.0 501B A23B at 67.6%. 5 models have a LongBench v2 result in our records.
Scores as the model makers published them, not re-run by us; each row links to its announcement.
Best reported score
70.6%
T Thinking Machines Inkling 2 Small
Reported by Thinking Machines, November 9, 2026
5 models ranked · 4 makers
| Rank | Model | Score | Setting | Reported by | Date |
|---|---|---|---|---|---|
| 1 | T Inkling 2 Small | 70.6% | Not published | Thinking Machines | November 9, 2026 (model release) |
| 2 | R Beam 1.0 501B A23B | 67.6% | FP8 | Reflection | October 16, 2026 (model release) |
| 3 | Qwen3.8 Max (0902) | 66.3% | Not published | Qwen | August 3, 2026 |
| 4 | Qwen3.7 Max | 65.3% | Not published | Qwen | August 3, 2026 |
| 5 | Kimi K2.5 | 61.0% | Thinking | Moonshot AI | January 27, 2026 (model release) |
Other reported runs
| Model | Score | Setting | Reported by | Date |
|---|---|---|---|---|
| R Beam 1.0 501B A23B | 67.5% | BF16 | Reflection | October 16, 2026 (model release) |
| R Beam 1.0 501B A23B | 67.2% | NVFP4 | Reflection | October 16, 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
- 70.6% to 61.0% across 5 models.
- Freshness
- Newest result published November 9, 2026; oldest January 27, 2026.
What LongBench v2 measures
Models answer 503 multiple-choice questions using contexts of 8,000 to 2 million words; the score reflects how many answers are correct. Source: THUDM/LongBench: LongBench v2 and LongBench