AutoLogi
Models solve open-ended logic puzzles in English and Chinese. Generated verifier programs check each answer against the puzzle constraints.
In the authors’ words
While logical reasoning evaluation of Large Language Models (LLMs) has attracted significant attention, existing benchmarks predominantly rely on multiple-choice formats that are vulnerable to random guessing, leading to overestimated performance and substantial performance fluctuations. To obtain more accurate assessments of models' reasoning capabilities, we propose an automated method for synthesizing open-ended logic puzzles, and use it to develop a bilingual benchmark, AutoLogi. Our approach features program-based verification and controllable difficulty levels, enabling more reliable evaluation that better distinguishes models' reasoning abilities. Extensive evaluation of eight modern LLMs shows that AutoLogi can better reflect true model capabilities, with performance scores spanning from 35% to 73% compared to the narrower range of 21% to 37% on the source multiple-choice dataset. Beyond benchmark creation, this synthesis method can generate high-quality training data by incorporating program verifiers into the rejection sampling process, enabling systematic enhancement of LLMs' reasoning capabilities across diverse datasets.
- Level
- School Students, from grade school to high school
- Scored
- Checked automatically A program checks the result against an answer key, unit tests or the expected final state.
- Answers
- Short answer A number, word or short string matched to a key.
Kimi K2 leads with 89.5% among models released by Jul 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.
The newest model with a score here came out in Jul 2025, and nobody has reported a newer one. Among the 20 highest-ranked models on the capability index, these have no score yet:
- GPT-6 Astra
- Claude Opus 5.5
- GPT-6.1 Terra
- GPT-6.1 Sol
- and 16 more
- 1 Kimi K2 Jul 2025 89.5%
- 2
Qwen3-14B 89.2% - 3 Qwen3-8B 89.1%
- 4 Qwen3-235B-A22B 89%
- 5 Qwen3-30B-A3B 88.7%
- 6 Qwen3-4B 87.9%
- 7 Qwen3-32B 87.3%
- 8 Qwen3-1.7B 83.2%
| # | Model | Score | Setting | Date | Source |
|---|---|---|---|---|---|
| 1 | Kimi K2 Jul 2025 | 89.5% | Jul 2025 | Model card | |
| 2 | Qwen3-14B | 89.2% | May 2025 | Model card | |
| 3 | Qwen3-8B | 89.1% | May 2025 | Model card | |
| 4 | Qwen3-235B-A22B | 89% | Apr 2025 | Model card | |
| 5 | Qwen3-30B-A3B | 88.7% | May 2025 | Model card | |
| 6 | Qwen3-4B | 87.9% | May 2025 | Model card | |
| 7 | Qwen3-32B | 87.3% | May 2025 | Model card | |
| 8 | Qwen3-1.7B | 83.2% | May 2025 | Model card |
3 models have held number one since its release in Feb 2025
Lab claims count from the day the card came out, independent runs from the model’s release.
The best score passed 85% in Apr 2025
All the scores we have, one dot each, so a model tested at several settings appears more than once.
Labs dropped it from launch posts once the best score passed 85%
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 |
|---|---|---|---|
| 11 Jul 2025 | Kimi K2 | 89.5 | AutoLogi, Acc |
| 29 Apr 2025 | Qwen3-235B-A22B | 89 | AutoLogi |
8 system cards report it, from 2 labs
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 |
|---|---|---|---|---|
| 21 Jul 2025 | Kimi K2 technical report | 89.5 | AutoLogi (Acc.) (Table 3) | Capabilities |
| 14 May 2025 | Qwen3-8B technical report | 89.1 | AutoLogi, thinking mode (Table 17). Non-thinking 76.5 (Table 18) | Capabilities |
| 14 May 2025 | Qwen3-4B technical report | 87.9 | AutoLogi, thinking mode (Table 17). Non-thinking 76.3 (Table 18) | Capabilities |
| 14 May 2025 | Qwen3-32B technical report | 87.3 | AutoLogi, thinking mode (Table 13). Non-thinking 78.5 (Table 14) | Capabilities |
| 14 May 2025 | Qwen3-30B-A3B technical report | 88.7 | AutoLogi, thinking mode (Table 15). Non-thinking 81.5 (Table 16) | Capabilities |
| 14 May 2025 | Qwen3-235B-A22B technical report | 89 | AutoLogi, thinking mode (Table 11). Non-thinking 83.3 (Table 12) | Capabilities |
| 14 May 2025 | Qwen3-14B technical report | 89.2 | AutoLogi, thinking mode (Table 15). Non-thinking 82.0 (Table 16) | Capabilities |
| 14 May 2025 | Qwen3-1.7B technical report | 83.2 | AutoLogi, thinking mode (Table 19). Non-thinking 59.8 (Table 20) | Capabilities |
More in Science and reasoning
- GPQA Diamond In use 198 graduate-level 'Google-proof' multiple-choice questions in biology, chemistry and physics.
Gemini 4 Argon 96.4% - Terminal-Bench-Science In use 70 scientific research-workflow tasks across five science areas, run in Harbor containers. Hidden tests give the resolution rate. GPT-6 Astra 68.1%
- FrontierScience In use Models solve physics, chemistry, and biology problems in an Olympiad track graded by short answer and a Research track graded by 10-point rubrics.
Intern-S2-397B 87% - BioMysteryBench In use Bioinformatics problems on anonymized real datasets that require real analysis with tools. Rubric-graded, with human-solvable and human-difficult splits. Claude Opus 5 90.1%
- HiPhO In use Models solve theory problems from 2024 and 2025 physics olympiad exams, some with diagrams. Official marking schemes grade answers and steps, mapped to medal levels. Gemini 3 Pro 78.3%
- LABBench2 In use 11 biology research sub-tasks (literature, figures, tables, patents, trials and more) with tool access. Macro-averaged accuracy. GPT-5.2 60.9%