Benchmark Almanac

Chess LLM Benchmark (Ryan Brew)

Gemma 4 31B IT 35.8% leads since April 2026
Not in model cards

Models evaluate chess positions, choose a move, and explain the position. The primary move metric counts blunders against the Lc0 chess engine.

Brew ranks 22 models on 4,000 Lichess positions. The pooled blunder rate was 41.9% among legal moves, including moves made after a retry.

Made by
Ryan Brew
Kind
Puzzles and chess
Level
School
Scored
Checked automatically
Answers
Short answer
Tasks
All public
Measures
Capability
Language
English
In and out
Text → Text
Interaction
Single turn
Items from
Real work and use
Refresh
One fixed set
Run by
Anyone
Strongest model tested
Gemma 4 31B IT #101 on the capability index
Last model added
April 2026
Scores refreshed
November 20th, 2026
22
Entries on its own leaderboard, 18 models we track

Gemma 4 31B IT leads with 35.8% among models released by April 2026 Gemma 4 31B IT leads with 35.8% among models released by April 2026

Best score per entry from the benchmark's own leaderboard, read on November 20th, 2026. Where a model ran with several settings or harnesses, we keep its best. Names in gray aren't in our model list: fine-tunes, agent systems, products and older models. They don't count toward leaders. This view shows only models we track, each at its best.

The newest model with a score here came out in April 2026, and nobody has tested a newer one. Among the 20 highest-ranked models on the capability index, these have no score yet:

  1. 1 Gemma 4 31B IT 35.8%
  2. 2 Gemma 4 26B A4B 37.3%
  3. 3 Mistral 7B v0.1 39.7%
  4. 4 solar:10.7b 40.4%
  5. 5 Gemma 3 12B 41%
  6. 6 Llama 3.2 3B 41.1%
  7. 7 Mixtral 8x7B 41.5%
  8. 8 DeepSeek-R1-Distill-Qwen-7B 41.6%
  9. 9 codellama:34b 42%
  10. 10 Qwen2.5-7B 42.1%
  11. 11 Qwen2.5-14B 42.3%
  12. 12 command-r:35b 42.5%
  13. 13 Yi-34B 42.5%
  14. 14 Gemma 4 E4B 42.6%
  15. 15 DeepSeek-R1-Distill-Qwen-14B 42.6%
  16. 16 Phi-4 43%
  17. 17 Gemma 3 4B 43%
  18. 18 Llama 3.3 70B 43.2%
  19. 19 wizardlm2:7b 43.5%
  20. 20 Qwen2.5-32B 44.2%
  21. 21 Llama 3.1-8B 44.4%
  22. 22 Gemma 4 E2B 44.5%

Source: Official Chess LLM Benchmark (Ryan Brew) leaderboard

Source: Official Chess LLM Benchmark (Ryan Brew) leaderboard

#EntryScore ±SettingDateSource
1 Gemma 4 31B IT 35.8% ±0.7 Apr 2026 Official leaderboard
2 Gemma 4 26B A4B 37.3% ±0.8 Apr 2026 Official leaderboard
3 Mistral 7B v0.1 39.7% ±0.8 Sep 2023 Official leaderboard
4 solar:10.7b 40.4% ±0.8 Official leaderboard
5 Gemma 3 12B 41% ±0.8 Mar 2025 Official leaderboard
6 Llama 3.2 3B 41.1% ±0.8 Sep 2024 Official leaderboard
7 Mixtral 8x7B 41.5% ±0.8 Dec 2023 Official leaderboard
8 DeepSeek-R1-Distill-Qwen-7B 41.6% ±0.8 Jan 2025 Official leaderboard
9 codellama:34b 42% ±0.8 Official leaderboard
10 Qwen2.5-7B 42.1% ±0.8 Sep 2024 Official leaderboard
11 Qwen2.5-14B 42.3% ±0.8 Sep 2024 Official leaderboard
12 command-r:35b 42.5% ±0.8 Official leaderboard
13 Yi-34B 42.5% ±0.8 Nov 2023 Official leaderboard
14 Gemma 4 E4B 42.6% ±0.8 Mar 2026 Official leaderboard
15 DeepSeek-R1-Distill-Qwen-14B 42.6% ±0.8 Jan 2025 Official leaderboard
16 Phi-4 43% ±0.8 Dec 2024 Official leaderboard
17 Gemma 3 4B 43% ±0.8 Mar 2025 Official leaderboard
18 Llama 3.3 70B 43.2% ±0.8 Dec 2024 Official leaderboard
19 wizardlm2:7b 43.5% ±0.8 Official leaderboard
20 Qwen2.5-32B 44.2% ±0.8 Sep 2024 Official leaderboard
21 Llama 3.1-8B 44.4% ±0.8 Jul 2024 Official leaderboard
22 Gemma 4 E2B 44.5% ±0.8 Mar 2026 Official leaderboard
#ModelScore ±SettingDateSource
1 Gemma 4 31B IT 35.8% ±0.7 Apr 2026 Official leaderboard
2 Gemma 4 26B A4B 37.3% ±0.8 Apr 2026 Official leaderboard
3 Mistral 7B v0.1 39.7% ±0.8 Sep 2023 Official leaderboard
4 Gemma 3 12B 41% ±0.8 Mar 2025 Official leaderboard
5 Llama 3.2 3B 41.1% ±0.8 Sep 2024 Official leaderboard
6 Mixtral 8x7B 41.5% ±0.8 Dec 2023 Official leaderboard
7 DeepSeek-R1-Distill-Qwen-7B 41.6% ±0.8 Jan 2025 Official leaderboard
8 Qwen2.5-7B 42.1% ±0.8 Sep 2024 Official leaderboard
9 Qwen2.5-14B 42.3% ±0.8 Sep 2024 Official leaderboard
10 Yi-34B 42.5% ±0.8 Nov 2023 Official leaderboard
11 Gemma 4 E4B 42.6% ±0.8 Mar 2026 Official leaderboard
12 DeepSeek-R1-Distill-Qwen-14B 42.6% ±0.8 Jan 2025 Official leaderboard
13 Phi-4 43% ±0.8 Dec 2024 Official leaderboard
14 Gemma 3 4B 43% ±0.8 Mar 2025 Official leaderboard
15 Llama 3.3 70B 43.2% ±0.8 Dec 2024 Official leaderboard
16 Qwen2.5-32B 44.2% ±0.8 Sep 2024 Official leaderboard
17 Llama 3.1-8B 44.4% ±0.8 Jul 2024 Official leaderboard
18 Gemma 4 E2B 44.5% ±0.8 Mar 2026 Official leaderboard

2 models have held number one on the official leaderboard since 2023

Lab claims count from the day the card came out, independent runs from the model’s release.

Mistral AI Google DeepMind
Own leaderboard Not tested since Apr 2026

Source: Official Chess LLM Benchmark (Ryan Brew) leaderboard

Scores over time

All the scores we have, one dot each, so a model tested at several settings appears more than once.

Plot with other benchmarks
On the benchmark’s own leaderboard Best so far
0% 20% 40% 60% 80% 100% 2024 2025 2026 Gemma 4 31B IT Mistral 7B v0.1 0% 20% 40% 60% 80% 100% 2024 2025 2026 Gemma 4 31B IT Mistral 7B v0.1

Source: Official Chess LLM Benchmark (Ryan Brew) 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 Games and long tasks

All 261 in Games and long tasks
  • ARC-AGI-3 In use Interactive game environments where agents must explore, infer goals and rules, and act efficiently without instructions. GPT-6.2 Astra 100%
  • Orak Fading LLM agents play 12 popular video games across most genres through an MCP interface. Each game scores in-game progress or outcomes, and the board ranks models by their average rank across the games. Gemini 2.5 Pro 3.90
  • MineDojo Fading Agents complete open-ended tasks in Minecraft, such as mining blocks and fighting mobs, from a large simulation suite. Programmatic checks score many tasks. Seed2.0 Pro 49%
  • TextQuests Only in system cards Agents play 25 classic interactive fiction games. The score is harmful actions per 100 steps, where lower is better. The board also reports game progress. Llama 4 Scout 7.00
  • GameWorld Only in system cards Multimodal game agents play 34 browser games across 170 tasks. Outcomes are checked from the game state. GPT-6.1 Terra 76.8%
  • Swarm Bench Only in system cards Models orchestrate agent swarms on complex tasks that benefit from coordinated decomposition and parallel execution. Kimi K3 76.3%

Data audit

8 of 9 checks pass, 1 warning. Look first at the Newest score check. The newest came 7 months ago, in April 2026, for Gemma 4 26B A4B. Every benchmark

Catalog record 2 pass
  • Pass Task access All tasks are public.
  • Pass Confidence Checked against a primary source.
Scores 1 warning 2 pass 2 notes
  • Pass Independent runs Scores from the official leaderboard (22 entries).
  • Pass Models scored 18 models we track, plus 4 other entries on its leaderboard.
  • Warning Newest score The newest came 7 months ago, in April 2026, for Gemma 4 26B A4B.
  • Note Frontier models None of the top 10 has a score. The strongest tested is Gemma 4 31B IT (#101).
  • Note Headroom Lower is better, so there’s no ceiling to reach.
Leaderboard import 3 pass
  • Pass Importer We read the board on github.com: 18 rows for models we track and 4 other entries 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.

How we check