Benchmark Almanac

BoardBench 0.2

GPT-6 Astra 67.2 leads since October 2026
Not in model cards
Version 0.2

Models choose one move in each of 30 board positions across ten games. An engine grades each legal move and assigns zero to illegal moves.

The maker ranks eight models. GPT-6 Astra scored 67.2 out of 100 and Gemini 4 Argon scored 66.3.

Made by
BoardBench
Released
October 2026
Kind
Puzzles and chess
Level
Everyday
Scored
Checked automatically
Answers
Short answer
Tasks
All public
Measures
Capability
Language
English
In and out
Text → Text
Interaction
Single turn
Run by
Anyone
Based on
BoardBench 0.1
Strongest model tested
GPT-6 Astra #4 on the capability index
Last model added
October 2026
Scores refreshed
November 20th, 2026
8
Models scored on its own leaderboard

GPT-6 Astra leads with 67.2

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. Every run is in the next chart.

  1. 1 GPT-6 Astra high 67.2
  2. 2 Gemini 4 Argon high 66.3
  3. 3 GPT-6.1 Sol high 56.4
  4. 4 GLM-5.3-Flash high 43.9
  5. 5 GPT-6 Luna high 42.5
  6. 6 DeepSeek V4.1 Flash high 39.9
  7. 7 Claude Haiku 5.5 high 39.0
  8. 8 Claude Opus 5.5 high 37.6

Source: Official BoardBench 0.2 leaderboard

#ModelScoreSettingDateLinkSource
1 GPT-6 Astra 67.2 high Sep 2026 boardbench.dev Official leaderboard
2 Gemini 4 Argon 66.3 high Sep 2026 boardbench.dev Official leaderboard
3 GPT-6.1 Sol 56.4 high Sep 2026 boardbench.dev Official leaderboard
4 GLM-5.3-Flash 43.9 high Aug 2026 boardbench.dev Official leaderboard
5 GPT-6 Luna 42.5 high Sep 2026 boardbench.dev Official leaderboard
6 DeepSeek V4.1 Flash 39.9 high Sep 2026 boardbench.dev Official leaderboard
7 Claude Haiku 5.5 39.0 high Oct 2026 boardbench.dev Official leaderboard
8 Claude Opus 5.5 37.6 high Sep 2026 boardbench.dev Official leaderboard

Rankings by part

We read 10 parts from the benchmark’s own leaderboard, each ranking models on one part of the test, such as Xiangqi, Chess and Go. GPT-6 Astra leads 5 of them, and 4 other models lead the rest. Arrows show how far a model moved from its overall rank among the models we track.

Xiangqi

8 models we track have a score here.

  1. 1 GPT-6 Astra 88.3
  2. 2 GPT-6.1 Sol ↑1 88.3
  3. 3 Gemini 4 Argon ↓1 55.2
  4. 4 GLM-5.3-Flash 55.0
  5. 5 Claude Opus 5.5 ↑3 55.0
  6. 6 DeepSeek V4.1 Flash 33.3
  7. 7 GPT-6 Luna ↓2 21.9
  8. 8 Claude Haiku 5.5 ↓1 0.00

The best score, 67.2, costs $0.56 a task

Runs on the cost frontier score higher than any run that's cheaper. Gemini 4 Argon at high comes within 0.9 of the best for $0.28 a task, about half the cost. The costs are the leaderboard's own estimates, so compare them only within this benchmark.

Z.ai OpenAI Google DeepMind Anthropic Cost frontier One model at each effort level
0.00 20.0 40.0 60.0 80.0 $0.001 $0.01 $0.10 $1.00 Cost per task, log scale GPT-6 Astra, high GPT-6.1 Sol, high GLM-5.3-Flash, high GPT-6 Luna, high Claude Haiku 5.5, high 0.00 20.0 40.0 60.0 80.0 $0.001 $0.01 $0.10 $1.00 Cost per task, log scale GPT-6 Astra, high GPT-6.1 Sol, high

Source: Official BoardBench 0.2 leaderboard

ModelEffortHarnessScoreCost per taskOutput tokens
GPT-6 Astra high 67.2 $0.56
Gemini 4 Argon high 66.3 $0.28
GPT-6.1 Sol high 56.4 $0.10
GLM-5.3-Flash high 43.9 $0.01
GPT-6 Luna high 42.5 $0.004
DeepSeek V4.1 Flash high 39.9 $0.03
Claude Haiku 5.5 high 39.0 $0.003
Claude Opus 5.5 high 37.6 $0.08

Number one over time

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

OpenAI
Own leaderboard 1 leader

Source: Official BoardBench 0.2 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
30.0 40.0 50.0 60.0 70.0 30.0 40.0 50.0 60.0 70.0

Source: Official BoardBench 0.2 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.

Versions

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 (8 entries).
  • Pass Models scored 8 models we track.
  • Pass Newest score The newest is from October 2026, for Claude Haiku 5.5.
  • Pass Frontier models 4 of the top 10 have a score, led by GPT-6 Astra (#4).
  • Note Headroom Scored as a plain number, with no fixed ceiling.
Leaderboard import 3 pass
  • Pass Importer We read the board on boardbench.dev: 8 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