LongCoT
Models solve long reasoning problems across five domains. Task-specific programs verify answers, with optional judge fallback for ambiguous mathematics and chemistry responses.
The April 2026 paper reports GPT-5.2 leading with 9.83% accuracy on full LongCoT, followed by Gemini 3 Pro.
- Made by
- Sumeet Ramesh Motwani and colleagues
- Released
- April 2026
- Kind
- Reasoning over long documents
- Level
- Expert
- Scored
- Checked automatically
- Answers
- Short answer
- Tasks
- All public
- Measures
- Capability
- Language
- English
- Interaction
- Single turn
- Refresh
- One fixed set
- Run by
- Anyone
- In views
- Chemistry and materials
- Strongest model tested
- GPT-5.2 #51 on the capability index
- Last model added
- February 2026
- Scores refreshed
- November 20th, 2026
GPT-5.2 leads with 25.1% among models released by February 2026
Best score per model from the benchmark's own leaderboard, read on November 20th, 2026. Where a model ran with several settings or harnesses, we keep its best.
The newest model with a score here came out in February 2026, and nobody has tested a newer one. Among the 20 highest-ranked models on the capability index, these have no score yet:
- GPT-6.2 Astra
- Claude Fable 5.2
-
Gemini 4 Carbon
- GPT-6 Astra
- and 16 more
- 1 GPT-5.2 Open Harness ✓ 25.1%
- 2
Qwen3.5-27B Open Harness 22.2%
- 3
Qwen3.5-9B Open Harness 15.6%
- 4
Gemini 3 Pro Raw LLM ✓ 6.08%
- 5 Grok 4.1 Fast Raw LLM ✓ 2.04%
- 6 Claude Sonnet 4.5 Raw LLM ✓ 1.86%
- 7
DeepSeek-V3.2 Raw LLM ✓ 1.52%
- 8 Kimi K2 Thinking Raw LLM ✓ 1.24%
- 9 GLM-4.7 Raw LLM ✓ 0.48%
Source: Official LongCoT leaderboard
| # | Model | Score | Setting | Submitted by | Date | Link | Source |
|---|---|---|---|---|---|---|---|
| 1 | GPT-5.2 ✓ | 25.1% | Open Harness, rlm | LongCoT Team | Dec 2025 | github.com | Official leaderboard |
| 2 | | 22.2% | Open Harness, dspy.RLM | Raymond Weitekamp | Feb 2026 | dspy.ai | Official leaderboard |
| 3 | | 15.6% | Open Harness, dspy.RLM | Raymond Weitekamp | Feb 2026 | dspy.ai | Official leaderboard |
| 4 | | 6.08% | Raw LLM | LongCoT Team | Nov 2025 | Official leaderboard | |
| 5 | Grok 4.1 Fast ✓ | 2.04% | Raw LLM | LongCoT Team | Nov 2025 | Official leaderboard | |
| 6 | Claude Sonnet 4.5 ✓ | 1.86% | Raw LLM | LongCoT Team | Sep 2025 | Official leaderboard | |
| 7 | | 1.52% | Raw LLM | LongCoT Team | Dec 2025 | Official leaderboard | |
| 8 | Kimi K2 Thinking ✓ | 1.24% | Raw LLM | LongCoT Team | Nov 2025 | Official leaderboard | |
| 9 | GLM-4.7 ✓ | 0.48% | Raw LLM | LongCoT Team | Dec 2025 | Official leaderboard |
Rankings by part
We read 5 parts from the benchmark’s own leaderboard, each ranking models on one part of the test, such as Logic, CS and Chemistry. Qwen3.5-27B leads 2 of them, and 2 other models lead the rest. Arrows show how far a model moved from its overall rank among the models we track.
Logic
9 models we track have a score here.
- 1
Qwen3.5-27B ↑1 71.6%
- 2 GPT-5.2 ↓1 68.3%
- 3
Qwen3.5-9B 56.7%
- 4
Gemini 3 Pro 1.50%
- 5 Claude Sonnet 4.5 ↑1 0.50%
- 6 Kimi K2 Thinking ↑2 0.50%
- 7
DeepSeek-V3.2 0.30%
- 8 Grok 4.1 Fast ↓3 0%
- 9 GLM-4.7 0%
CS
9 models we track have a score here.
- 1 GPT-5.2 26.7%
- 2
Qwen3.5-27B 4.20%
- 3
Gemini 3 Pro ↑1 4.10%
- 4 Grok 4.1 Fast ↑1 3.10%
- 5 Claude Sonnet 4.5 ↑1 2.20%
- 6
DeepSeek-V3.2 ↑1 1.50%
- 7 Kimi K2 Thinking ↑1 0.70%
- 8 GLM-4.7 ↑1 0.40%
- 9
Qwen3.5-9B ↓6 0%
Chemistry
9 models we track have a score here.
- 1
Gemini 3 Pro ↑3 12.5%
- 2 GPT-5.2 ↓1 10.1%
- 3 Claude Sonnet 4.5 ↑3 2.10%
- 4 Grok 4.1 Fast ↑1 1.80%
- 5
DeepSeek-V3.2 ↑2 0.60%
- 6 Kimi K2 Thinking ↑2 0.30%
- 7 GLM-4.7 ↑2 0.30%
- 8
Qwen3.5-27B ↓6 0%
- 9
Qwen3.5-9B ↓6 0%
Chess
9 models we track have a score here.
- 1
Qwen3.5-27B ↑1 35%
- 2 GPT-5.2 ↓1 30.6%
- 3
Qwen3.5-9B 21.8%
- 4
Gemini 3 Pro 4.90%
- 5 Claude Sonnet 4.5 ↑1 2.30%
- 6 GLM-4.7 ↑3 1.70%
- 7 Kimi K2 Thinking ↑1 1.40%
- 8
DeepSeek-V3.2 ↓1 1.40%
- 9 Grok 4.1 Fast ↓4 0%
Math
9 models we track have a score here.
- 1 GPT-5.2 9.80%
- 2
Gemini 3 Pro ↑2 7.50%
- 3 Grok 4.1 Fast ↑2 5.50%
- 4
DeepSeek-V3.2 ↑3 3.80%
- 5 Kimi K2 Thinking ↑3 3.30%
- 6 Claude Sonnet 4.5 2.20%
- 7 GLM-4.7 ↑2 0%
- 8
Qwen3.5-27B ↓6 0%
- 9
Qwen3.5-9B ↓6 0%
Number one over time
Lab claims count from the day the card came out, independent runs from the model’s release.
Source: Official LongCoT leaderboard
Scores over time
All the scores we have, one dot each, so a model tested at several settings appears more than once.
Source: Official LongCoT leaderboard
No lab model card reports it yet
We read the headline results of 540 launch posts from 57 labs, and none of them list this benchmark.
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%
Data audit
9 of 10 checks pass, 1 warning. Look first at the Newest score check. The newest came 9 months ago, in February 2026, for Qwen3.5-27B. Every benchmark
Catalog record 2 pass
- Pass Task access All tasks are public.
- Pass Confidence Checked against a primary source.
Scores 1 warning 3 pass 1 note
- Pass Independent runs Scores from the official leaderboard (9 entries).
- Pass Models scored 9 models we track.
- Warning Newest score The newest came 9 months ago, in February 2026, for Qwen3.5-27B.
- Note Frontier models None of the top 10 has a score. The strongest tested is GPT-5.2 (#51).
- Pass Headroom The best is 25.1%, under 85%, so it still separates models.
Leaderboard import 3 pass
- Pass Importer We read the board on longcot.ai: 11 rows for models we track at the last read.
- Pass Last read November 20th, 2026.
- Pass Read worked The last read worked.
Consistency 1 pass 1 note
- Note Lab claims No model has both a lab number and an independent run.
- Pass Known issues No published audit found problems.