Benchmark Almanac

SettlerBench Season One

GPT-6 Astra 1,270 leads since October 2026
Not in model cards

Four models trade resources, build settlements, and negotiate in public or private messages. Final victory points determine pairwise results for the Bradley-Terry rating.

Synopticon ranks 21 models across 357 games. Its 24 rating rows distinguish hosting and reasoning settings for two models.

Released
October 2026
Kind
Multi-agent games
Level
Everyday
Scored
Ranked by matches
Answers
Agent task
Tasks
All public
Measures
Capability
Language
English
In and out
Text → Text, Actions
Interaction
Multi-agent
Environment
Catan
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
September 2026
Scores refreshed
November 20th, 2026
21
Models scored on its own leaderboard

2 models share the top score of 1,270 among models released by September 2026

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.

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

Bars start at 400.

  1. 1 GPT-6 Astra 1,270
  2. 2 GPT-6.1 Sol 1,270
  3. 3 Gemini 3.8 Flash 1,213
  4. 4 Gemini 3.7 Flash 1,201
  5. 5 Kimi K3 1,190
  6. 6 Claude Fable 5.1 1,190
  7. 7 Grok 4.7 1,178
  8. 8 Claude Opus 5.5 1,148
  9. 9 GLM-5.3-Flash full thinking 1,102
  10. 10 Muse Spark 1.3 1,091
  11. 11 DeepSeek-V4-Pro Novita host 1,063
  12. 12 DeepSeek V4.1 Flash 1,059
  13. 13 Gemma 4 31B IT 962
  14. 14 Qwen 3.8 Max 959
  15. 15 GPT-6 Luna 922
  16. 16 Qwen 3.8 27B 848
  17. 17 MiniMax-M3 820
  18. 18 GPT-5.6 Luna 758
  19. 19 Muse Glimmer 726
  20. 20 gpt-oss-120b 585
  21. 21 Nemotron 3.5 Lightning 30B-A3B 553

Source: Official SettlerBench Season One leaderboard

#ModelScore ±SettingDateSource
1 GPT-6 Astra 1,270 1200 to 1349 Sep 2026 Official leaderboard
2 GPT-6.1 Sol 1,270 ±85 Sep 2026 Official leaderboard
3 Gemini 3.8 Flash 1,213 1146 to 1299 Sep 2026 Official leaderboard
4 Gemini 3.7 Flash 1,201 1142 to 1269 Aug 2026 Official leaderboard
5 Kimi K3 1,190 1116 to 1271 Jul 2026 Official leaderboard
6 Claude Fable 5.1 1,190 1120 to 1271 Sep 2026 Official leaderboard
7 Grok 4.7 1,178 1096 to 1266 Sep 2026 Official leaderboard
8 Claude Opus 5.5 1,148 1030 to 1258 Sep 2026 Official leaderboard
9 GLM-5.3-Flash 1,102 ±81 full thinking Aug 2026 Official leaderboard
10 Muse Spark 1.3 1,091 1035 to 1152 Sep 2026 Official leaderboard
11 DeepSeek-V4-Pro 1,063 ±76.5 Novita host Apr 2026 Official leaderboard
12 DeepSeek V4.1 Flash 1,059 ±54.5 Sep 2026 Official leaderboard
13 Gemma 4 31B IT 962 ±68 Apr 2026 Official leaderboard
14 Qwen 3.8 Max 959 898 to 1017 Aug 2026 Official leaderboard
15 GPT-6 Luna 922 ±92.5 Sep 2026 Official leaderboard
16 Qwen 3.8 27B 848 ±67 Aug 2026 Official leaderboard
17 MiniMax-M3 820 751 to 884 Jun 2026 Official leaderboard
18 GPT-5.6 Luna 758 681 to 828 Jul 2026 Official leaderboard
19 Muse Glimmer 726 653 to 793 Aug 2026 Official leaderboard
20 gpt-oss-120b 585 508 to 657 Aug 2025 Official leaderboard
21 Nemotron 3.5 Lightning 30B-A3B 553 469 to 617 Aug 2026 Official leaderboard

The best score, 1,270, costs $1.46 a task

Runs on the cost frontier score higher than any run that's cheaper. The costs are the leaderboard's own estimates, so compare them only within this benchmark.

Google DeepMind DeepSeek OpenAI Cost frontier One model at each effort level
0 500 1,000 1,500 $0.01 $0.10 $1.00 $10 Cost per task, log scale GPT-6.1 Sol Gemini 3.8 Flash DeepSeek V4.1 Flash Gemma 4 31B IT GPT-6 Luna 0 500 1,000 1,500 $0.01 $0.10 $1.00 $10 Cost per task, log scale GPT-6.1 Sol DeepSeek V4.1 Flash

Source: Official SettlerBench Season One leaderboard

ModelEffortHarnessScoreCost per taskOutput tokens
GPT-6 Astra 1,270 $6.43
GPT-6.1 Sol 1,270 $1.46
Gemini 3.8 Flash 1,213 $0.74
Gemini 3.7 Flash 1,201 $0.62
Kimi K3 1,190 $9.83
Claude Fable 5.1 1,190 $8.97
Grok 4.7 1,178 $2.85
Claude Opus 5.5 1,148 $2.82
GLM-5.3-Flash full thinking 1,102 $0.86
Muse Spark 1.3 1,091 $1.73
DeepSeek-V4-Pro Novita host 1,063 $2.86
DeepSeek V4.1 Flash 1,059 $0.61
DeepSeek-V4-Pro Ionstream host 1,006 $1.38
DeepSeek-V4-Pro mixed hosts 994 $1.96
Gemma 4 31B IT 962 $0.32
Qwen 3.8 Max 959 $2.44
GPT-6 Luna 922 $0.09
GLM-5.3-Flash low thinking 892 $0.31
Qwen 3.8 27B 848 $3.06
MiniMax-M3 820 $1.06
GPT-5.6 Luna 758 $0.19
Muse Glimmer 726 $0.50
gpt-oss-120b 585 $0.11
Nemotron 3.5 Lightning 30B-A3B 553 $0.35

Number one over time

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

OpenAI
Own leaderboard Not tested since Oct 2026
GPT-6 Astra 1,270

Source: Official SettlerBench Season One 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
400 600 800 1,000 1,200 1,400 2026 400 600 800 1,000 1,200 1,400 2026

Source: Official SettlerBench Season One 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 (21 entries).
  • Pass Models scored 21 models we track.
  • Pass Newest score The newest is from September 2026, for GPT-6.1 Sol.
  • Pass Frontier models 4 of the top 10 have a score, led by GPT-6 Astra (#4).
  • Note Headroom Scored as a rating, with no fixed ceiling.
Leaderboard import 3 pass
  • Pass Importer We read the board on synopticonresearch.com: 24 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