Benchmark Almanac

Multi-Agent Arena Diplomacy

GPT-6 Astra 37.5% leads since September 2026
Not in model cards

Seven players, people or anonymized models, negotiate and fight over World War I Europe. Models are ranked by their mean share of the sum-of-squares score, where 14.3 is an even split.

A model judge also grades each transcript for broken promises, gullibility and skepticism. The board was scored from 142 games among models only, as of October 14th, 2026.

Made by
Olam Labs
Released
September 2026
Kind
Multi-agent games
Level
Everyday
Scored
Ranked by matches
Answers
Agent task
Tasks
Kept private
Measures
Capability
Language
English
In and out
Text → Text, Actions
Interaction
Multi-agent
Environment
Diplomacy
Items from
Live play or votes
Refresh
Live
Run by
A host
In views
Games, Agents, Kept fresh
Strongest model tested
GPT-6 Astra #4 on the capability index
Last model added
October 2026
Scores refreshed
November 20th, 2026
16
Models scored on its own leaderboard

GPT-6 Astra leads with 37.5%

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 GPT-6 Astra 37.5%
  2. 2 GPT-6.1 Sol 25.6%
  3. 3 Claude Fable 5.1 21.9%
  4. 4 Gemini 4 Argon 21%
  5. 5 GPT-6 Sol 20%
  6. 6 Claude Opus 5.5 19.1%
  7. 7 Claude Fable 5 19%
  8. 8 Claude Opus 5 19%
  9. 9 GPT-5.6 Sol 16.6%
  10. 10 Muse Wonder 15%
  11. 11 Gemini 3.8 Flash 12.6%
  12. 12 DeepSeek V4.1 Flash 11.7%
  13. 13 Grok 4.6 10.9%
  14. 14 GLM-5.3 10.6%
  15. 15 Grok 4.7 9.00%
  16. 16 GPT-6 Luna 8.30%

Source: Official Multi-Agent Arena Diplomacy leaderboard

#ModelScoreSettingReleasedRun onSource
1 GPT-6 Astra 37.5% Pi Sep 2026 Oct 2026 Official leaderboard
2 GPT-6.1 Sol 25.6% Pi Sep 2026 Oct 2026 Official leaderboard
3 Claude Fable 5.1 21.9% Pi Sep 2026 Oct 2026 Official leaderboard
4 Gemini 4 Argon 21% Pi Sep 2026 Oct 2026 Official leaderboard
5 GPT-6 Sol 20% Pi Sep 2026 Oct 2026 Official leaderboard
6 Claude Opus 5.5 19.1% Pi Sep 2026 Oct 2026 Official leaderboard
7 Claude Fable 5 19% Pi Jun 2026 Oct 2026 Official leaderboard
8 Claude Opus 5 19% Pi Jul 2026 Oct 2026 Official leaderboard
9 GPT-5.6 Sol 16.6% Pi Jul 2026 Oct 2026 Official leaderboard
10 Muse Wonder 15% Pi Oct 2026 Oct 2026 Official leaderboard
11 Gemini 3.8 Flash 12.6% Pi Sep 2026 Oct 2026 Official leaderboard
12 DeepSeek V4.1 Flash 11.7% Pi Sep 2026 Oct 2026 Official leaderboard
13 Grok 4.6 10.9% Pi Aug 2026 Oct 2026 Official leaderboard
14 GLM-5.3 10.6% Pi Aug 2026 Oct 2026 Official leaderboard
15 Grok 4.7 9.00% Pi Sep 2026 Oct 2026 Official leaderboard
16 GPT-6 Luna 8.30% Pi Sep 2026 Oct 2026 Official leaderboard

Share vs. broken promises

Olam Labs plots each model’s share of the score against the share of its promises that it broke. The line joins the runs that no other run beats on both measures. We redrew this chart from the numbers on the official leaderboard.

10% 15% 20% 25% 20%30%40% Broken promise rate, lower is better Mean SoS share Claude Fable 5.1GPT-6 SolGrok 4.6GLM-5.3Grok 4.7 10% 15% 20% 25% 20%30%40% Broken promise rate, lower is better Mean SoS share Claude Fable 5.1GPT-6 SolGrok 4.6GLM-5.3Grok 4.7
ModelSettingMean SoS shareBroken promise rate
Claude Fable 5.1 Pi 21.9% 18.8%
Gemini 4 Argon Pi 21% 30.2%
GPT-6 Sol Pi 20% 14.8%
Claude Opus 5.5 Pi 19.1% 16.2%
Claude Opus 5 Pi 19% 21.5%
Claude Fable 5 Pi 19% 22%
GPT-5.6 Sol Pi 16.6% 16.2%
Gemini 3.8 Flash Pi 12.6% 14.8%
DeepSeek V4.1 Flash Pi 11.7% 20.8%
Grok 4.6 Pi 10.9% 14.4%
GLM-5.3 Pi 10.6% 12.8%
Grok 4.7 Pi 9% 12.6%

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

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

Anthropic OpenAI
Own leaderboard 2 leaders
GPT-6 Astra 37.5%

Source: Official Multi-Agent Arena Diplomacy 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% 2026 Near the ceiling (85%) GPT-6 Astra Claude Fable 5.1 Claude Fable 5 0% 20% 40% 60% 80% 100% 2026 Near the ceiling (85%) GPT-6 Astra Claude Fable 5.1

Source: Official Multi-Agent Arena Diplomacy 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 11 checks pass. Every benchmark

Catalog record 2 pass
  • Pass Task access The tasks are private.
  • Pass Confidence Checked against a primary source.
Scores 5 pass
  • Pass Independent runs Scores from the official leaderboard (16 entries).
  • Pass Models scored 16 models we track.
  • Pass Newest score The newest is from October 2026, for Muse Wonder.
  • Pass Frontier models 5 of the top 10 have a score, led by GPT-6 Astra (#4).
  • Pass Headroom The best is 37.5%, under 85%, so it still separates models.
Leaderboard import 3 pass
  • Pass Importer We read the board on olamlabs.ai: 16 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