Benchmark
LongBench v2
What is LongBench v2?
LongBench v2 is a benchmark designed to assess the ability of LLMs to handle long-context problems requiring deep understanding and reasoning across real-world…
- Released
- 2024-12-19
- Modality
- text
- Evaluates
- long context, reasoning, structured output, general
- Openness
- unknown
- Importer
- LLM Stats
- Reported scores
- 17
Source provenance
- Original evidence https://api.zeroeval.com/leaderboard/benchmarks/longbench-v2?top_n=500
LongBench v2 results and reported scores
LLM Stats
| Model | Organization | Reported value | Reported | Evidence |
|---|---|---|---|---|
| DeepSeek-V3 | DeepSeek | 0.487 | 2024-12-25 | evidence |
| Kimi K2.5 | Moonshot AI | 0.61 | 2026-01-27 | evidence |
| MAI-Thinking-1 | Microsoft | 0.61 | 2026-06-02 | evidence |
| MiMo-V2-Flash | Xiaomi | 0.606 | 2025-12-16 | evidence |
| MiniMax M1 40K | MiniMax | 0.61 | 2025-06-16 | evidence |
| MiniMax M1 80K | MiniMax | 0.615 | 2025-06-16 | evidence |
| Nemotron 3 Ultra (550B A55B) | NVIDIA | 0.619 | 2026-06-04 | evidence |
| Qwen3.5-0.8B | Qwen | 0.261 | 2026-03-02 | evidence |
| Qwen3.5-122B-A10B | Qwen | 0.602 | 2026-02-24 | evidence |
| Qwen3.5-27B | Qwen | 0.606 | 2026-02-24 | evidence |
| Qwen3.5-2B | Qwen | 0.387 | 2026-03-02 | evidence |
| Qwen3.5-35B-A3B | Qwen | 0.59 | 2026-02-24 | evidence |
| Qwen3.5-397B-A17B | Qwen | 0.632 | 2026-02-16 | evidence |
| Qwen3.5-4B | Qwen | 0.5 | 2026-03-02 | evidence |
| Qwen3.5-9B | Qwen | 0.552 | 2026-03-02 | evidence |
| Qwen3.6 Plus | Qwen | 0.62 | 2026-04-02 | evidence |
| Qwen3.8 Max | Qwen | 0.663 | 2026-08-02 | evidence |
Scores are partitioned by the source that reported them and are never merged into a single cross-source ranking, because the sources measure different things and say so.