LongBench-Chat
Qwen2.5-14B-Instruct-1M 8.76 leads since January 2025 Qwen2.5-14B-Instruct-1M 8.76 leads since January 2025 Models follow real user instructions over contexts of 10K to 100K tokens. GPT-4 scores each answer from 1 to 10 against references.
Introduced with the LongAlign paper.
- Made by
- Tsinghua University and Zhipu AI (Bai et al., LongAlign)
- Released
- January 2024
- Kind
- Reasoning over long documents
- Level
- Everyday
- Scored
- Graded by a model
- Answers
- Open-ended
- Tasks
- All public
- Strongest model tested
- Qwen2.5-72B #165 on the capability index
- Last in a launch post
- January 2025, dropped before it saturated
- Last model added
- January 2025
- Scores refreshed
- November 12th, 2026
Qwen2.5-14B-Instruct-1M leads with 8.76 among models released by January 2025
Best score per model as labs report them in their own cards. Labs pick their own settings, so these compare less cleanly than independent runs.
Claude 2.1 leads with 8.03 across all independent runs
Best score per model across the official leaderboard, Epoch AI’s own runs and any board Epoch AI copied. Each bar and row names its source.
Qwen2.5-14B-Instruct-1M leads with 8.76, claims included
Best score per model from every source, including labs’ own numbers and claims Epoch AI collected from reports and papers. Labs pick their own settings, so a claim can beat an independent run of the same model.
The newest model with a score here came out in January 2025, and nobody has reported a newer one. Among the 20 highest-ranked models on the capability index, these have no score yet:
- Claude Fable 5.2
-
Gemini 4 Carbon - GPT-6 Astra
- Claude Opus 5.5
- and 16 more
- 1 Qwen2.5-14B-Instruct-1M 8.76
- 2 Qwen2.5-72B 8.72
- 3 Qwen2.5-32B 8.70
- 4 Qwen-Turbo 8.34
- 5 Qwen2.5-14B 8.04
- 6 Qwen2.5-7B 7.42
Source: Labs’ model cards
Source: Official LongBench-Chat leaderboard
- 1 Qwen2.5-14B-Instruct-1M 8.76
- 2 Qwen2.5-72B 8.72
- 3 Qwen2.5-32B 8.70
- 4 Qwen-Turbo 8.34
- 5 Qwen2.5-14B 8.04
- 6 Claude 2.1 8.03
- 7 Qwen2.5-7B 7.42
- 8
Mistral-7B-Instruct-v0.2 5.84
Sources: Labs’ model cards and official LongBench-Chat leaderboard
| # | Model | Score | Setting | Date | Source |
|---|---|---|---|---|---|
| 1 | Qwen2.5-14B-Instruct-1M | 8.76 | Jan 2025 | Lab model card | |
| 2 | Qwen2.5-72B | 8.72 | Dec 2024 | Lab model card | |
| 3 | Qwen2.5-32B | 8.70 | Dec 2024 | Lab model card | |
| 4 | Qwen-Turbo | 8.34 | Dec 2024 | Lab model card | |
| 5 | Qwen2.5-14B | 8.04 | Dec 2024 | Lab model card | |
| 6 | Qwen2.5-7B | 7.42 | Dec 2024 | Lab model card |
| # | Model | Score | Setting | Date | Source |
|---|---|---|---|---|---|
| 1 | Claude 2.1 | 8.03 | Nov 2023 | Official leaderboard | |
| 2 | Mistral-7B-Instruct-v0.2 | 5.84 | Dec 2023 | Official leaderboard |
| # | Model | Score | Setting | Date | Source |
|---|---|---|---|---|---|
| 1 | Qwen2.5-14B-Instruct-1M | 8.76 | Jan 2025 | Lab model card | |
| 2 | Qwen2.5-72B | 8.72 | Dec 2024 | Lab model card | |
| 3 | Qwen2.5-32B | 8.70 | Dec 2024 | Lab model card | |
| 4 | Qwen-Turbo | 8.34 | Dec 2024 | Lab model card | |
| 5 | Qwen2.5-14B | 8.04 | Dec 2024 | Lab model card | |
| 6 | Claude 2.1 | 8.03 | Nov 2023 | Official leaderboard | |
| 7 | Qwen2.5-7B | 7.42 | Dec 2024 | Lab model card | |
| 8 | Mistral-7B-Instruct-v0.2 | 5.84 | Dec 2023 | Official leaderboard |
The benchmark's own leaderboard has 14 entries, 2 of them models we track. Switch the scores above to rank them together.
2 models have held number one in labs’ own numbers since its release in January 2024 Number one over time 3 models have held number one with claims included since its release in January 2024
Lab claims count from the day the card came out, independent runs from the model’s release.
Source: Labs’ model cards
Scores over time
All the scores we have, one dot each, so a model tested at several settings appears more than once.
Sources: Official LongBench-Chat leaderboard and labs’ model cards
No launch post has listed it since January 2025
Launch posts from every lab that has listed it, one tick per post. Tall colored ticks include it, and short gray ones leave it out.
| Date | Launch post | Score | As the post writes it |
|---|---|---|---|
| Jan 27, 2025 | Qwen2.5-14B-Instruct-1M | 8.76 | LongBench-Chat |
6 system cards report it, from 1 lab
Besides launch posts, we read the system cards, model cards and technical reports labs publish with a release. They report far more, including safety and frontier-risk tests.
| Date | Document | Score | As the document writes it | Section |
|---|---|---|---|---|
| Jan 26, 2025 | Qwen2.5-14B-Instruct-1M technical report | 8.76 | LongBench-Chat (Table 5, Table 3). 8.56 before RL | Capabilities |
| Dec 19, 2024 | Qwen2.5-7B technical report | 7.42 | LongBench-Chat (Table 17) | Capabilities |
| Dec 19, 2024 | Qwen2.5-72B technical report | 8.72 | LongBench-Chat (Table 17) | Capabilities |
| Dec 19, 2024 | Qwen2.5-32B technical report | 8.7 | LongBench-Chat (Table 17) | Capabilities |
| Dec 19, 2024 | Qwen2.5-14B technical report | 8.04 | LongBench-Chat (Table 17) | Capabilities |
| Dec 19, 2024 | Qwen-Turbo technical report | 8.34 | LongBench-Chat (Table 17) | Capabilities |
More in Long context
- MRCR v2 In use Multi-round coreference: retrieve the n-th of several similar requests hidden in a long synthetic conversation, up to 1M tokens.
Muse Spark 1.3 98.5% - LongBench v2 In use 503 multiple-choice questions over 8K-2M-word contexts (QA, dialogue, code repos, structured data). Accuracy.
Inkling 2 Small 70.6% - GraphWalks In use A graph is given as an edge list in long context. The model answers BFS or parent queries, scored by F1. Claude Fable 5 99.96%
- OpenAI MRCR v2 In use Models read a long synthetic chat that hides 2, 4 or 8 identical writing requests and must reproduce the response to the i-th one. Scored by mean string match ratio per context bin. Gemini 3.7 Flash 89%
- CL-bench In use Model must learn new knowledge from a supplied long context (rules, procedures, data) and solve tasks. Strict LLM-verified rubrics. GPT-5.4 27.9%
- AA-LCR 1.1 In use Models answer 100 questions that need reasoning across sets of real documents of about 100K tokens. An LLM equality checker with a fixed system prompt compares each answer with the key.
Step 5 Preview 88.3%