Benchmark Almanac

The Aggregate Lichess Rating Prediction

Gemini 3.1 Pro 270 leads since February 2026
Not in model cards

Models estimate both players' pregame Lichess rapid ratings from the moves of 25 anonymized games. The score is mean absolute error across 50 players.

The maker ranks 19 models. Gemini 3.1 Pro Preview averaged 269.66 rating points of error. Always guessing 1500 gave 542.90 points of error.

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
GPT-6.2 Astra #1 on the capability index
Last model added
November 2026
Scores refreshed
November 20th, 2026
19
Models scored on its own leaderboard

Gemini 3.1 Pro leads with 270

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.

  1. 1 Gemini 3.1 Pro medium 270
  2. 2 Gemini 3 Flash medium 272
  3. 3 Gemini 3.8 Flash medium 312
  4. 4 Gemini 3.7 Flash medium 314
  5. 5 Claude Fable 5.2 medium 333
  6. 6 GPT-6 Astra medium 334
  7. 7 GPT-6.1 Sol medium 335
  8. 8 Gemini 3.6 Flash medium 339
  9. 9 Gemini 3.5 Flash medium 347
  10. 10 GPT-6.2 Astra medium 354
  11. 11 GPT-5.6 Sol medium 372
  12. 12 GPT-6.2 Sol medium 380
  13. 13 GPT-6 Sol medium 400
  14. 14 Claude Opus 5.5 medium 438
  15. 15 Claude Fable 5.1 medium 442
  16. 16 GLM-5.3-Flash medium 464
  17. 17 GPT-5.6 Luna medium 465
  18. 18 Qwen 3.8 Max prime, medium 481
  19. 19 Claude Sonnet 5.5 medium 492

Source: Official The Aggregate Lichess Rating Prediction leaderboard

#ModelScore ±SettingReleasedRun onSource
1 Gemini 3.1 Pro 270 199 to 344.6 medium Feb 2026 Oct 2026 Official leaderboard
2 Gemini 3 Flash 272 206.7 to 342.6 medium Dec 2025 Oct 2026 Official leaderboard
3 Gemini 3.8 Flash 312 225.9 to 404.4 medium Sep 2026 Oct 2026 Official leaderboard
4 Gemini 3.7 Flash 314 240 to 391.8 medium Aug 2026 Oct 2026 Official leaderboard
5 Claude Fable 5.2 333 248.9 to 424.7 medium Nov 2026 Nov 2026 Official leaderboard
6 GPT-6 Astra 334 244.5 to 430.7 medium Sep 2026 Oct 2026 Official leaderboard
7 GPT-6.1 Sol 335 239.4 to 440.9 medium Sep 2026 Oct 2026 Official leaderboard
8 Gemini 3.6 Flash 339 251.9 to 432.9 medium Jul 2026 Oct 2026 Official leaderboard
9 Gemini 3.5 Flash 347 ±92.5 medium May 2026 Oct 2026 Official leaderboard
10 GPT-6.2 Astra 354 265.6 to 445.9 medium Nov 2026 Nov 2026 Official leaderboard
11 GPT-5.6 Sol 372 274.9 to 475.3 medium Jul 2026 Oct 2026 Official leaderboard
12 GPT-6.2 Sol 380 ±92 medium Nov 2026 Nov 2026 Official leaderboard
13 GPT-6 Sol 400 291 to 513.6 medium Sep 2026 Oct 2026 Official leaderboard
14 Claude Opus 5.5 438 328.3 to 558.9 medium Sep 2026 Oct 2026 Official leaderboard
15 Claude Fable 5.1 442 326.3 to 573.2 medium Sep 2026 Oct 2026 Official leaderboard
16 GLM-5.3-Flash 464 358.5 to 579.4 medium Aug 2026 Oct 2026 Official leaderboard
17 GPT-5.6 Luna 465 ±112.1 medium Jul 2026 Oct 2026 Official leaderboard
18 Qwen 3.8 Max 481 351.6 to 625.6 prime, medium Aug 2026 Oct 2026 Official leaderboard
19 Claude Sonnet 5.5 492 373.2 to 619.1 medium Sep 2026 Oct 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.

Google DeepMind
Own leaderboard 2 leaders

Source: Official The Aggregate Lichess Rating Prediction 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
250 300 350 400 450 500 2026 2027 250 300 350 400 450 500 2026 2027

Source: Official The Aggregate Lichess Rating Prediction 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

All 10 checks pass. Every benchmark

Catalog record 2 pass
  • Pass Task access All tasks are public.
  • Pass Confidence Checked against a primary source.
Scores 4 pass 1 note
  • Pass Independent runs Scores from the official leaderboard (19 entries).
  • Pass Models scored 19 models we track.
  • Pass Newest score The newest is from November 2026, for GPT-6.2 Astra.
  • Pass Frontier models 7 of the top 10 have a score, led by GPT-6.2 Astra (#1).
  • Note Headroom Scored as a rating, with no fixed ceiling.
Leaderboard import 3 pass
  • Pass Importer We read the board on theaggregate.ai: 19 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.

How we check