Benchmark Almanac

RedstoneBench

GPT-6.2 Astra 88.7 leads since November 2026
Not in model cards

Agents design redstone circuits in Minecraft. A game server tests whether each circuit works and measures the space it occupies.

Adam Holter maintains a public leaderboard with replayable builds for 10 models, including GPT-6.2 Astra and Gemini 4 Neon.

Made by
Adam Holter
Released
November 2026
Kind
Video games
Level
Expert
Scored
Checked automatically
Answers
Agent task
Tasks
All public
Measures
Capability
Language
English
In and out
Text → Actions
Interaction
Agentic
Environment
Minecraft
Refresh
One fixed set
Run by
Anyone
In views
Games, Agents
Strongest model tested
GPT-6.2 Astra #1 on the capability index
Last model added
November 2026
Scores refreshed
November 20th, 2026
10
Models scored on its own leaderboard

GPT-6.2 Astra leads with 88.7

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-6.2 Astra max 88.7
  2. 2 GPT-6.2 Sol max 75.8
  3. 3 GPT-6 Astra ultra 63.9
  4. 4 Gemini 4 Neon high 63.4
  5. 5 GPT-6.1 Sol ultra 62.5
  6. 6 Gemini 4 Carbon high 59.4
  7. 7 Claude Opus 5.5 medium 57.1
  8. 8 GPT-6.1 Terra max 49.5
  9. 9 Grok 4.7 high 29.2
  10. 10 GPT-6 Luna max 19.2

Source: Official RedstoneBench leaderboard

#ModelScoreSettingDateSource
1 GPT-6.2 Astra 88.7 max Nov 2026 Official leaderboard
2 GPT-6.2 Sol 75.8 max Nov 2026 Official leaderboard
3 GPT-6 Astra 63.9 ultra Sep 2026 Official leaderboard
4 Gemini 4 Neon 63.4 high Nov 2026 Official leaderboard
5 GPT-6.1 Sol 62.5 ultra Sep 2026 Official leaderboard
6 Gemini 4 Carbon 59.4 high Oct 2026 Official leaderboard
7 Claude Opus 5.5 57.1 medium Sep 2026 Official leaderboard
8 GPT-6.1 Terra 49.5 max Oct 2026 Official leaderboard
9 Grok 4.7 29.2 high Sep 2026 Official leaderboard
10 GPT-6 Luna 19.2 max Sep 2026 Official leaderboard

Rankings by part

We read 16 parts from the benchmark’s own leaderboard in 5 groups: doors, piston extenders, logic, memory and timing. Each ranks models on one part of the test. GPT-6.2 Astra leads 10 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.

2x2 piston door

10 models we track have a score here.

  1. 1 GPT-6.1 Terra ↑7 100.0
  2. 2 Gemini 4 Carbon ↑4 100.0
  3. 3 GPT-6.2 Astra ↓2 100.0
  4. 4 GPT-6.2 Sol ↓2 100.0
  5. 5 Gemini 4 Neon ↓1 100.0
  6. 6 GPT-6 Astra ↓3 88.9
  7. 7 Claude Opus 5.5 88.9
  8. 8 GPT-6.1 Sol ↓3 88.9
  9. 9 Grok 4.7 66.7
  10. 10 GPT-6 Luna 66.7

More effort adds a median 6 score

Models the leaderboard ran at two or more effort levels. We don’t have enough cost data here to say what the extra effort costs.

Lowest effort to highest, light to dark
GPT-6.2 Astra
ultra low medium high max
88.7
GPT-6.2 Sol
ultra low medium max
75.8
GPT-6 Astra
ultra low high
63.9
Gemini 4 Neon
high
63.4
GPT-6.1 Sol
ultra low medium xhigh
62.5
Claude Opus 5.5
low medium high max
57.1
GPT-6.1 Terra
ultra medium max
49.5
Grok 4.7
low medium
29.2
GPT-6 Luna
low xhigh
19.2

Source: Official RedstoneBench leaderboard

ModelEffortHarnessScoreCost of the full runOutput tokens
GPT-6.2 Astra max 88.7
GPT-6.2 Astra ultra 87.6
GPT-6.2 Astra xhigh 82.0
GPT-6.2 Astra high 81.5
GPT-6.2 Sol max 75.8
GPT-6.2 Astra medium 72.1
GPT-6.2 Sol ultra 69.2
GPT-6.2 Sol high 64.3
GPT-6 Astra ultra 63.9
Gemini 4 Neon high 63.4
GPT-6.1 Sol ultra 62.5
GPT-6.2 Astra low 61.4
GPT-6.2 Sol xhigh 60.9
Gemini 4 Neon xhigh 60.3
GPT-6 Astra max 59.8
GPT-6 Astra xhigh 59.7
Gemini 4 Carbon high 59.4
GPT-6.1 Sol xhigh 59.3
GPT-6.2 Sol medium 58.0
GPT-6.1 Sol max 57.3
Claude Opus 5.5 medium 57.1
GPT-6 Astra high 54.4
GPT-6.1 Sol high 50.3
GPT-6.1 Terra max 49.5
Claude Opus 5.5 high 49.5
Claude Opus 5.5 xhigh 48.6
GPT-6.1 Sol medium 48.1
GPT-6.2 Sol low 48.1
GPT-6 Astra medium 46.0
GPT-6 Astra low 45.1
Claude Opus 5.5 low 43.7
GPT-6.1 Terra ultra 39.4
GPT-6.1 Sol low 37.5
GPT-6.1 Terra xhigh 36.4
GPT-6.1 Terra high 33.8
GPT-6.1 Terra low 32.6
GPT-6.1 Terra medium 31.9
Grok 4.7 high 29.2
Claude Opus 5.5 max 28.1
Grok 4.7 medium 27.2
Grok 4.7 xhigh 21.6
GPT-6 Luna max 19.2
GPT-6 Luna xhigh 17.8
Grok 4.7 low 16.1
GPT-6 Luna high 8.47
GPT-6 Luna medium 7.74
GPT-6 Luna low 4.17

2 models have held number one on the official leaderboard since its release in November 2026

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

OpenAI
Own leaderboard 2 leaders

Source: Official RedstoneBench 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
0.00 20.0 40.0 60.0 80.0 100.0 2027 0.00 20.0 40.0 60.0 80.0 100.0 2027

Source: Official RedstoneBench 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 (10 entries).
  • Pass Models scored 10 models we track.
  • Pass Newest score The newest is from November 2026, for Gemini 4 Neon.
  • Pass Frontier models 6 of the top 10 have a score, led by GPT-6.2 Astra (#1).
  • Note Headroom Scored as a plain number, with no fixed ceiling.
Leaderboard import 3 pass
  • Pass Importer We read the board on redstone-bench.adamholter.chatgpt.site: 47 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