Benchmark Almanac

BAZAAR

o3 7 leads since April 2025
Not in model cards

Agents buy and sell goods in a simulated competitive market.

The author publishes the model comparison in the repository.

Made by
Lech Mazur
Kind
Multi-agent games
Level
University
Scored
Settled by events
Answers
Agent task
Tasks
All public
Measures
Capability
Language
English
In and out
Text → Actions
Interaction
Multi-agent
Environment
A simulator
Refresh
Live
Run by
Anyone
Strongest model tested
o3 #84 on the capability index
Last model added
September 2025
Scores refreshed
November 20th, 2026
53
Entries on its own leaderboard, 24 models we track

o3 leads with 7 among models released by September 2025 o3 leads with 7 among models released by September 2025

Best score per entry from the benchmark's own leaderboard, read on November 20th, 2026. Where a model ran with several settings or harnesses, we keep its best. Names in gray aren't in our model list: fine-tunes, agent systems, products and older models. They don't count toward leaders. This view shows only models we track, each at its best.

The newest model with a score here came out in September 2025, 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 2.

  1. 1 o3 medium reasoning 7
  2. 2 Gemini 2.5 Pro Jun 2025 7
  3. 3 Gemini 2.5 Flash Sep 2025 7
  4. 4 Claude Sonnet 4 no reasoning 7
  5. 5 Grok 4 7
  6. 6 Claude Opus 4 Thinking 16K 7
  7. 7 o4-mini medium reasoning 7
  8. 8 DeepSeek-R1 May 2025 7
  9. 9 Qwen3-235B-A22B 7
  10. 10 GPT-4o Mar 2025 6
  11. 11 Enhanced Bayesian (Baseline) 6
  12. 12 Momentum (Baseline) 6
  13. 13 Threat Tit-for-Tat (Baseline) 6
  14. 14 Stochastic Shading Bandit (Baseline) 6
  15. 15 Contrarian (Baseline) 6
  16. 16 Llama 4 Maverick 6
  17. 17 Regime Switcher (Baseline) 5
  18. 18 Sniper (Baseline) 5
  19. 19 GDX (Baseline) 5
  20. 20 Risk-Aware (Baseline) 5
  21. 21 Regression (Baseline) 5
  22. 22 Signal Jammer (Baseline) 5
  23. 23 ZIC+ (Baseline) 5
  24. 24 Roth-Erev (Baseline) 5
  25. 25 Shading +5 (Baseline) 5
  26. 26 Adaptive Aggressive (Baseline) 5
  27. 27 Amazon Nova Pro 5
  28. 28 DeepSeek-V3 Mar 2025 5
  29. 29 Mean Reversion (Baseline) 5
  30. 30 Adaptive Aggressive Cliff (Baseline) 5
  31. 31 Shading +10 (Baseline) 5
  32. 32 Collusive (Baseline) 5
  33. 33 Truthful (Baseline) 5
  34. 34 Claude 3.5 Haiku 5
  35. 35 Qwen3-30B-A3B 5
  36. 36 Q-Learning (Baseline) 5
  37. 37 Llama 4 Scout 5
  38. 38 GPT-4o mini 5
  39. 39 ZIC Pure Biased Activity (Baseline) 5
  40. 40 Mistral Small 3.2 5
  41. 41 Adversarial Exploiter (Baseline) 5
  42. 42 Mistral Medium 3 5
  43. 43 Fictitious Play (Baseline) 5
  44. 44 MiniMax-Text-01 5
  45. 45 Gjerstad-Dickhaut (Baseline) 4
  46. 46 Phi-4 4
  47. 47 Kimi K2 Sep 2025 4
  48. 48 Penny Jumper (Baseline) 4
  49. 49 ERNIE-4.5-300B-A47B 4
  50. 50 Gemma 3 27B 4
  51. 51 LinearEq (Baseline) 4
  52. 52 Adaptive (Baseline) 2
  53. 53 Random (Baseline) 2

Source: Official BAZAAR leaderboard

#EntryScore ±SettingDateSource
1 o3 7 ±0.2 medium reasoning Apr 2025 Official leaderboard
2 Gemini 2.5 Pro Jun 2025 7 ±0.3 Jun 2025 Official leaderboard
3 Gemini 2.5 Flash Sep 2025 7 ±0.2 Sep 2025 Official leaderboard
4 Claude Sonnet 4 7 ±0.3 no reasoning May 2025 Official leaderboard
5 Grok 4 7 ±0.2 Jul 2025 Official leaderboard
6 Claude Opus 4 7 ±0.3 Thinking 16K May 2025 Official leaderboard
7 o4-mini 7 ±0.3 medium reasoning Apr 2025 Official leaderboard
8 DeepSeek-R1 May 2025 7 ±0.3 May 2025 Official leaderboard
9 Qwen3-235B-A22B 7 ±0.3 Apr 2025 Official leaderboard
10 GPT-4o Mar 2025 6 ±0.3 Mar 2025 Official leaderboard
11 Enhanced Bayesian (Baseline) 6 ±0 Official leaderboard
12 Momentum (Baseline) 6 ±0 Official leaderboard
13 Threat Tit-for-Tat (Baseline) 6 ±0 Official leaderboard
14 Stochastic Shading Bandit (Baseline) 6 ±0 Official leaderboard
15 Contrarian (Baseline) 6 ±0 Official leaderboard
16 Llama 4 Maverick 6 ±0.3 Apr 2025 Official leaderboard
17 Regime Switcher (Baseline) 5 ±0 Official leaderboard
18 Sniper (Baseline) 5 ±0 Official leaderboard
19 GDX (Baseline) 5 ±0 Official leaderboard
20 Risk-Aware (Baseline) 5 ±0 Official leaderboard
21 Regression (Baseline) 5 ±0 Official leaderboard
22 Signal Jammer (Baseline) 5 ±0 Official leaderboard
23 ZIC+ (Baseline) 5 ±0 Official leaderboard
24 Roth-Erev (Baseline) 5 ±0 Official leaderboard
25 Shading +5 (Baseline) 5 ±0 Official leaderboard
26 Adaptive Aggressive (Baseline) 5 ±0 Official leaderboard
27 Amazon Nova Pro 5 ±0.3 Dec 2024 Official leaderboard
28 DeepSeek-V3 Mar 2025 5 ±0.3 Mar 2025 Official leaderboard
29 Mean Reversion (Baseline) 5 ±0 Official leaderboard
30 Adaptive Aggressive Cliff (Baseline) 5 ±0 Official leaderboard
31 Shading +10 (Baseline) 5 ±0 Official leaderboard
32 Collusive (Baseline) 5 ±0 Official leaderboard
33 Truthful (Baseline) 5 ±0 Official leaderboard
34 Claude 3.5 Haiku 5 ±0.3 Oct 2024 Official leaderboard
35 Qwen3-30B-A3B 5 ±0.3 Apr 2025 Official leaderboard
36 Q-Learning (Baseline) 5 ±0 Official leaderboard
37 Llama 4 Scout 5 ±0.3 Apr 2025 Official leaderboard
38 GPT-4o mini 5 ±0.3 Jul 2024 Official leaderboard
39 ZIC Pure Biased Activity (Baseline) 5 ±0 Official leaderboard
40 Mistral Small 3.2 5 ±0.3 Jun 2025 Official leaderboard
41 Adversarial Exploiter (Baseline) 5 ±0 Official leaderboard
42 Mistral Medium 3 5 ±0.3 May 2025 Official leaderboard
43 Fictitious Play (Baseline) 5 ±0 Official leaderboard
44 MiniMax-Text-01 5 ±0.3 Jan 2025 Official leaderboard
45 Gjerstad-Dickhaut (Baseline) 4 ±0 Official leaderboard
46 Phi-4 4 ±0.3 Dec 2024 Official leaderboard
47 Kimi K2 Sep 2025 4 ±0.3 Sep 2025 Official leaderboard
48 Penny Jumper (Baseline) 4 ±0 Official leaderboard
49 ERNIE-4.5-300B-A47B 4 ±0.3 Jun 2025 Official leaderboard
50 Gemma 3 27B 4 ±0.4 Mar 2025 Official leaderboard
51 LinearEq (Baseline) 4 ±0 Official leaderboard
52 Adaptive (Baseline) 2 ±0 Official leaderboard
53 Random (Baseline) 2 ±0 Official leaderboard
#ModelScore ±SettingDateSource
1 o3 7 ±0.2 medium reasoning Apr 2025 Official leaderboard
2 Gemini 2.5 Pro Jun 2025 7 ±0.3 Jun 2025 Official leaderboard
3 Gemini 2.5 Flash Sep 2025 7 ±0.2 Sep 2025 Official leaderboard
4 Claude Sonnet 4 7 ±0.3 no reasoning May 2025 Official leaderboard
5 Grok 4 7 ±0.2 Jul 2025 Official leaderboard
6 Claude Opus 4 7 ±0.3 Thinking 16K May 2025 Official leaderboard
7 o4-mini 7 ±0.3 medium reasoning Apr 2025 Official leaderboard
8 DeepSeek-R1 May 2025 7 ±0.3 May 2025 Official leaderboard
9 Qwen3-235B-A22B 7 ±0.3 Apr 2025 Official leaderboard
10 GPT-4o Mar 2025 6 ±0.3 Mar 2025 Official leaderboard
11 Llama 4 Maverick 6 ±0.3 Apr 2025 Official leaderboard
12 Amazon Nova Pro 5 ±0.3 Dec 2024 Official leaderboard
13 DeepSeek-V3 Mar 2025 5 ±0.3 Mar 2025 Official leaderboard
14 Claude 3.5 Haiku 5 ±0.3 Oct 2024 Official leaderboard
15 Qwen3-30B-A3B 5 ±0.3 Apr 2025 Official leaderboard
16 Llama 4 Scout 5 ±0.3 Apr 2025 Official leaderboard
17 GPT-4o mini 5 ±0.3 Jul 2024 Official leaderboard
18 Mistral Small 3.2 5 ±0.3 Jun 2025 Official leaderboard
19 Mistral Medium 3 5 ±0.3 May 2025 Official leaderboard
20 MiniMax-Text-01 5 ±0.3 Jan 2025 Official leaderboard
21 Phi-4 4 ±0.3 Dec 2024 Official leaderboard
22 Kimi K2 Sep 2025 4 ±0.3 Sep 2025 Official leaderboard
23 ERNIE-4.5-300B-A47B 4 ±0.3 Jun 2025 Official leaderboard
24 Gemma 3 27B 4 ±0.4 Mar 2025 Official leaderboard

5 models have held number one on the official leaderboard since 2023

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

OpenAI Anthropic Amazon
Own leaderboard Not tested since Sep 2025
o3 7

Source: Official BAZAAR 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
4 5 6 7 8 2025 4 5 6 7 8 2025

Source: Official BAZAAR 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

8 of 9 checks pass, 1 warning. Look first at the Newest score check. The newest came 14 months ago, in September 2025, for Gemini 2.5 Flash. Every benchmark

Catalog record 2 pass
  • Pass Task access All tasks are public.
  • Pass Confidence Checked against a primary source.
Scores 1 warning 2 pass 2 notes
  • Pass Independent runs Scores from the official leaderboard (53 entries).
  • Pass Models scored 24 models we track, plus 29 other entries on its leaderboard.
  • Warning Newest score The newest came 14 months ago, in September 2025, for Gemini 2.5 Flash.
  • Note Frontier models None of the top 10 has a score. The strongest tested is o3 (#84).
  • Note Headroom Scored as a rating, with no fixed ceiling.
Leaderboard import 3 pass
  • Pass Importer We read the board on github.com: 26 rows for models we track and 29 other entries 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