Benchmark Almanac

PACT

GPT-5.5 1,607 leads since Apr 2026 Not in model cards

Two models bargain over 20 rounds as buyer and seller, each with a hidden private value. Matches across many pairings rank the models.

The repo was created on 2025-07-30 and updated on 2026-06-23.

Made by Lech Mazur Released Jul 2025 Multi-agent games All tasks public Last model added Jun 2026 Scores refreshed 2 Nov 2026
Level
Everyday Most adults, without training
Scored
Ranked by matches Models play against each other, and the results become ratings.
Answers
Open-ended Free-form text, documents, designs or media judged as a whole.
Code Leaderboard
26
Models scored on its own leaderboard

GPT-5.5 leads with 1,607 among models released by Jun 2026

Best score per model from the benchmark's own leaderboard, read on 2 Nov 2026. Where a model ran with several settings or harnesses, we keep its best.

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

26 models. Scroll sideways for the rest. Bars start at 1,350.

  1. 1,607 GPT-5.5
  2. 1,603 Claude Fable 5
  3. 1,570 DeepSeek-V4-Pro
  4. 1,566 GLM-5.2
  5. 1,562 Claude Opus 4.8
  6. 1,560 Kimi K2.6
  7. 1,557 Gemini 3.1 Pro
  8. 1,557 Gemma 4 31B IT
  9. 1,554 GLM-5.1
  10. 1,546 Claude Sonnet 4.6
  11. 1,538 Mistral Medium 3.5
  12. 1,526 Trinity Large Thinking
  13. 1,520 Tencent Hy3 preview
  14. 1,511 Qwen3.7-Max
  15. 1,510 Claude Haiku 4.5
  16. 1,509 Gemini 3.1 Flash-Lite
  17. 1,508 MiMo-V2.5-Pro
  18. 1,504 Grok 4.3
  19. 1,500 Step 3.7 Flash
  20. 1,496 gpt-oss-120b
  21. 1,491 Seed2.0 Pro
  22. 1,485 ERNIE 5.1
  23. 1,483 DeepSeek-V4-Flash
  24. 1,449 Mistral Large 3
  25. 1,426 Amazon Nova Pro
  26. 1,394 Llama 4 Maverick

Bars start at 1,350.

  1. 1 GPT-5.5 high 1,607
  2. 2 Claude Fable 5 high 1,603
  3. 3 DeepSeek-V4-Pro 1,570
  4. 4 GLM-5.2 max 1,566
  5. 5 Claude Opus 4.8 high 1,562
  6. 6 Kimi K2.6 1,560
  7. 7 Gemini 3.1 Pro 1,557
  8. 8 Gemma 4 31B IT reasoning 1,557
  9. 9 GLM-5.1 1,554
  10. 10 Claude Sonnet 4.6 high 1,546
  11. 11 Mistral Medium 3.5 high 1,538
  12. 12 Trinity Large Thinking 1,526
  13. 13 Tencent Hy3 preview 1,520
  14. 14 Qwen3.7-Max 1,511
  15. 15 Claude Haiku 4.5 1,510
  16. 16 Gemini 3.1 Flash-Lite 1,509
  17. 17 MiMo-V2.5-Pro 1,508
  18. 18 Grok 4.3 1,504
  19. 19 Step 3.7 Flash high 1,500
  20. 20 gpt-oss-120b 1,496
  21. 21 Seed2.0 Pro 1,491
  22. 22 ERNIE 5.1 1,485
  23. 23 DeepSeek-V4-Flash 1,483
  24. 24 Mistral Large 3 1,449
  25. 25 Amazon Nova Pro 1,426
  26. 26 Llama 4 Maverick 1,394
#ModelScoreSettingDateSource
1 GPT-5.5 1,607 high Apr 2026 Official leaderboard
2 Claude Fable 5 1,603 high Jun 2026 Official leaderboard
3 DeepSeek-V4-Pro 1,570 Apr 2026 Official leaderboard
4 GLM-5.2 1,566 max Jun 2026 Official leaderboard
5 Claude Opus 4.8 1,562 high May 2026 Official leaderboard
6 Kimi K2.6 1,560 Apr 2026 Official leaderboard
7 Gemini 3.1 Pro 1,557 Feb 2026 Official leaderboard
8 Gemma 4 31B IT 1,557 reasoning Apr 2026 Official leaderboard
9 GLM-5.1 1,554 Apr 2026 Official leaderboard
10 Claude Sonnet 4.6 1,546 high Feb 2026 Official leaderboard
11 Mistral Medium 3.5 1,538 high Apr 2026 Official leaderboard
12 Trinity Large Thinking 1,526 Apr 2026 Official leaderboard
13 Tencent Hy3 preview 1,520 Apr 2026 Official leaderboard
14 Qwen3.7-Max 1,511 May 2026 Official leaderboard
15 Claude Haiku 4.5 1,510 Oct 2025 Official leaderboard
16 Gemini 3.1 Flash-Lite 1,509 Mar 2026 Official leaderboard
17 MiMo-V2.5-Pro 1,508 Apr 2026 Official leaderboard
18 Grok 4.3 1,504 Apr 2026 Official leaderboard
19 Step 3.7 Flash 1,500 high May 2026 Official leaderboard
20 gpt-oss-120b 1,496 Aug 2025 Official leaderboard
21 Seed2.0 Pro 1,491 Feb 2026 Official leaderboard
22 ERNIE 5.1 1,485 May 2026 Official leaderboard
23 DeepSeek-V4-Flash 1,483 Apr 2026 Official leaderboard
24 Mistral Large 3 1,449 Dec 2025 Official leaderboard
25 Amazon Nova Pro 1,426 Dec 2024 Official leaderboard
26 Llama 4 Maverick 1,394 Apr 2025 Official leaderboard

7 models have held number one since its release in Jul 2025

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

Amazon OpenAI Anthropic Google DeepMind Moonshot AI
Own leaderboard Not tested since Jun 2026
GPT-5.5 1,607

Scores over time

All the scores we have, one dot each, so a model tested at several settings appears more than once.

On the benchmark’s own leaderboard
1,400 1,500 1,600 2025 2026 1,400 1,500 1,600 2025 2026

No lab model card reports it yet

We read the headline results of 479 launch posts from 51 labs, and none of them list this benchmark.

More in Games and long tasks

All 68 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 Astra 100%
  • 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%
  • Orak Fading LLM agents play 12 popular video games across most genres through an MCP interface. Scores come from in-game progress or outcomes per title. Nemotron 3 Nano 4B 22.9
  • GameWorld Only in system cards Multimodal game agents play 34 browser games across 170 tasks. Outcomes are checked from the game state. Gemini 3 Flash 41.9%
  • 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%
  • TextQuests Only in system cards Agents play 25 classic interactive fiction games. Measures game progress and harmful actions per 100 steps. Muse Spark 2.40