Benchmark Almanac

Agent Poker Bench

GPT-5.2 1,132 leads since Jan 2026 Not in model cards

Models play 10-seat no-limit Texas Hold'em tournaments against each other. Chip counts and elimination order feed TrueSkill ratings.

17 models. Not updated since 2025-12-23, when GPT 5.2 led at 1131.8.

Made by Vals AI (with The Graphite Group) Released Jan 2026 Multi-agent games Tasks kept private Last model added Dec 2025 Scores refreshed 2 Nov 2026
Level
Expert Specialists with years of training, often over hours of work
Scored
Ranked by matches Models play against each other, and the results become ratings.
Answers
Agent task Many steps in a terminal, browser, desktop, app, game or robot.
Leaderboard
17
Models scored on its own leaderboard

GPT-5.2 leads with 1,132 among models released by Dec 2025

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 Dec 2025, and nobody has tested a newer one. Among the 20 highest-ranked models on the capability index, these have no score yet:

17 models. Scroll sideways for the rest. Bars start at 850.

  1. 1,132 GPT-5.2
  2. 1,103 GPT-5
  3. 1,100 Gemini 3 Flash
  4. 1,090 DeepSeek-V3.2
  5. 1,079 Grok 4.1 Fast
  6. 1,079 Gemini 3 Pro
  7. 1,058 DeepSeek-V3.1
  8. 1,056 Claude Sonnet 4.5
  9. 1,039 GPT-5.1
  10. 1,034 Grok 4 Fast
  11. 1,033 Claude Opus 4.5
  12. 1,033 Gemini 2.5 Pro
  13. 1,015 gpt-oss-120b
  14. 1,012 Kimi K2 Thinking
  15. 995 Qwen3-Max
  16. 946 GLM-4.6
  17. 891 Llama 4 Maverick

Bars start at 850.

  1. 1 GPT-5.2 1,132
  2. 2 GPT-5 1,103
  3. 3 Gemini 3 Flash 1,100
  4. 4 DeepSeek-V3.2 thinking 1,090
  5. 5 Grok 4.1 Fast 1,079
  6. 6 Gemini 3 Pro 1,079
  7. 7 DeepSeek-V3.1 1,058
  8. 8 Claude Sonnet 4.5 non-thinking 1,056
  9. 9 GPT-5.1 1,039
  10. 10 Grok 4 Fast 1,034
  11. 11 Claude Opus 4.5 non-thinking 1,033
  12. 12 Gemini 2.5 Pro Jun 2025 1,033
  13. 13 gpt-oss-120b 1,015
  14. 14 Kimi K2 Thinking 1,012
  15. 15 Qwen3-Max 995
  16. 16 GLM-4.6 946
  17. 17 Llama 4 Maverick 891
#ModelScoreSettingDateSource
1 GPT-5.2 1,132 Dec 2025 Official leaderboard
2 GPT-5 1,103 Aug 2025 Official leaderboard
3 Gemini 3 Flash 1,100 Dec 2025 Official leaderboard
4 DeepSeek-V3.2 1,090 thinking Dec 2025 Official leaderboard
5 Grok 4.1 Fast 1,079 Nov 2025 Official leaderboard
6 Gemini 3 Pro 1,079 Nov 2025 Official leaderboard
7 DeepSeek-V3.1 1,058 Aug 2025 Official leaderboard
8 Claude Sonnet 4.5 1,056 non-thinking Sep 2025 Official leaderboard
9 GPT-5.1 1,039 Nov 2025 Official leaderboard
10 Grok 4 Fast 1,034 Sep 2025 Official leaderboard
11 Claude Opus 4.5 1,033 non-thinking Nov 2025 Official leaderboard
12 Gemini 2.5 Pro Jun 2025 1,033 Jun 2025 Official leaderboard
13 gpt-oss-120b 1,015 Aug 2025 Official leaderboard
14 Kimi K2 Thinking 1,012 Nov 2025 Official leaderboard
15 Qwen3-Max 995 Sep 2025 Official leaderboard
16 GLM-4.6 946 Sep 2025 Official leaderboard
17 Llama 4 Maverick 891 Apr 2025 Official leaderboard

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 Jan 2026
GPT-5.2 1,132

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
900 1,000 1,100 2025 2026 900 1,000 1,100 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