GTBench
Models play strategic games against other models or scripted opponents. The benchmark measures their advantage in game rewards.
The June 2024 paper compares GPT, Llama and DeepSeek models. GPT-4 led the reported model comparison with an average normalized reward advantage of 0.13.
- Made by
- GTBench authors
- Released
- February 2024
- Kind
- Multi-agent games
- Level
- Everyday
- Scored
- Ranked by matches
- Answers
- Agent task
- Tasks
- All public
- Measures
- Capability
- Language
- English
- Interaction
- Multi-agent
- Environment
- A simulator
- Items from
- Live play or votes
- Refresh
- Live
- Run by
- Anyone
- In views
- Agents, Robotics, Kept fresh
- Strongest model tested
- GPT-4 Jun 2023 #186 on the capability index
- Last model added
- January 2024
- Scores refreshed
- November 20th, 2026
GPT-4 leads with 0.13 among models released by January 2024 GPT-4 leads with 0.13 among models released by January 2024
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. Every run is in the next chart. 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 January 2024, and nobody has tested a newer one. Among the 20 highest-ranked models on the capability index, these have no score yet:
- GPT-6.2 Astra
- Claude Fable 5.2
-
Gemini 4 Carbon
- GPT-6 Astra
- and 16 more
- 1 GPT-4 Jun 2023, Opponent: GPT-3.5-turbo-1106 (prompt agent) 0.13
- 2 GPT-3.5 Turbo Jan 2024, Opponent: GPT-3.5-turbo-1106 (prompt agent) 0.02
- 3
Code Llama 34B Instruct Opponent: GPT-3.5-turbo-1106 (prompt agent) -0.01
- 4 Llama-2-70b-chat Opponent: GPT-3.5-turbo-1106 (prompt agent) -0.06
- 5 Mistral-7b-Orca Opponent: GPT-3.5-turbo-1106 (prompt agent) -0.11
Source: Official GTBench leaderboard
Source: Official GTBench leaderboard
| # | Entry | Score | Setting | Date | Source |
|---|---|---|---|---|---|
| 1 | GPT-4 Jun 2023 | 0.13 | Opponent: GPT-3.5-turbo-1106 (prompt agent), Prompt | Jun 2023 | Official leaderboard |
| 2 | GPT-3.5 Turbo Jan 2024 | 0.02 | Opponent: GPT-3.5-turbo-1106 (prompt agent), CoT | Jan 2024 | Official leaderboard |
| 3 | | -0.01 | Opponent: GPT-3.5-turbo-1106 (prompt agent), Prompt | Aug 2023 | Official leaderboard |
| 4 | Llama-2-70b-chat | -0.06 | Opponent: GPT-3.5-turbo-1106 (prompt agent), SC-CoT | Official leaderboard | |
| 5 | Mistral-7b-Orca | -0.11 | Opponent: GPT-3.5-turbo-1106 (prompt agent), CoT | Official leaderboard |
| # | Model | Score | Setting | Date | Source |
|---|---|---|---|---|---|
| 1 | GPT-4 Jun 2023 | 0.13 | Opponent: GPT-3.5-turbo-1106 (prompt agent), Prompt | Jun 2023 | Official leaderboard |
| 2 | GPT-3.5 Turbo Jan 2024 | 0.02 | Opponent: GPT-3.5-turbo-1106 (prompt agent), CoT | Jan 2024 | Official leaderboard |
| 3 | | -0.01 | Opponent: GPT-3.5-turbo-1106 (prompt agent), Prompt | Aug 2023 | Official leaderboard |
Rankings by part
We read 10 parts from the benchmark’s own leaderboard, each ranking models on one part of the test, such as Tic-Tac-Toe, Connect Four and Breakthrough. Code Llama 34B Instruct leads 6 of them, and 2 other models lead the rest. Arrows show how far a model moved from its overall rank among the models we track.
Tic-Tac-Toe
3 models we track have a score here.
- 1 GPT-3.5 Turbo Jan 2024 ↑1 0.41
- 2
Code Llama 34B Instruct ↑1 0.33
- 3 GPT-4 Jun 2023 ↓2 0.00
Connect Four
3 models we track have a score here.
- 1
Code Llama 34B Instruct ↑2 0.72
- 2 GPT-3.5 Turbo Jan 2024 0.24
- 3 GPT-4 Jun 2023 ↓2 0.08
Breakthrough
3 models we track have a score here.
- 1 GPT-4 Jun 2023 0.56
- 2 GPT-3.5 Turbo Jan 2024 0.16
- 3
Code Llama 34B Instruct -0.56
Liar's Dice
3 models we track have a score here.
- 1 GPT-4 Jun 2023 0.80
- 2 GPT-3.5 Turbo Jan 2024 0.52
- 3
Code Llama 34B Instruct -0.28
Blind Auction
3 models we track have a score here.
- 1
Code Llama 34B Instruct ↑2 0.45
- 2 GPT-3.5 Turbo Jan 2024 0.12
- 3 GPT-4 Jun 2023 ↓2 0.07
Negotiation
3 models we track have a score here.
- 1
Code Llama 34B Instruct ↑2 0.22
- 2 GPT-3.5 Turbo Jan 2024 0.18
- 3 GPT-4 Jun 2023 ↓2 0.14
Kuhn Poker
3 models we track have a score here.
- 1 GPT-4 Jun 2023 0.44
- 2 GPT-3.5 Turbo Jan 2024 0.12
- 3
Code Llama 34B Instruct 0.08
Nim
3 models we track have a score here.
- 1
Code Llama 34B Instruct ↑2 0.48
- 2 GPT-3.5 Turbo Jan 2024 0.12
- 3 GPT-4 Jun 2023 ↓2 0.08
Pig
3 models we track have a score here.
- 1
Code Llama 34B Instruct ↑2 0.12
- 2 GPT-4 Jun 2023 ↓1 0.04
- 3 GPT-3.5 Turbo Jan 2024 ↓1 0.00
Iterated Prisoner's Dilemma
3 models we track have a score here.
- 1
Code Llama 34B Instruct ↑2 0.60
- 2 GPT-3.5 Turbo Jan 2024 0.13
- 3 GPT-4 Jun 2023 ↓2 0.00
Scores at each effort level
Models the leaderboard ran at two or more effort levels. We don’t have enough cost data here to say what the extra effort costs.
Source: Official GTBench leaderboard
| Model | Effort | Harness | Score | Cost per task | Output tokens |
|---|---|---|---|---|---|
| GPT-4 Jun 2023 | Opponent: GPT-3.5-turbo-1106 (prompt agent) | Prompt | 0.13 | ||
| GPT-4 Jun 2023 | Opponent: GPT-3.5-turbo-1106 (prompt agent) | CoT | 0.13 | ||
| GPT-3.5 Turbo Jan 2024 | Opponent: GPT-3.5-turbo-1106 (prompt agent) | CoT | 0.02 | ||
| GPT-4 Jun 2023 | Opponent: GPT-4 (prompt agent) | Prompt | 0.00 | ||
| GPT-3.5 Turbo Jan 2024 | Opponent: GPT-3.5-turbo-1106 (prompt agent) | Prompt | 0.00 | ||
| GPT-3.5 Turbo Jan 2024 | Opponent: GPT-3.5-turbo-1106 (prompt agent) | SC-CoT | -0.00 | ||
| GPT-3.5 Turbo Jan 2024 | Opponent: GPT-3.5-turbo-1106 (prompt agent) | ToT | -0.01 | ||
| Opponent: GPT-3.5-turbo-1106 (prompt agent) | Prompt | -0.01 | |||
| Opponent: GPT-4 (prompt agent) | Prompt | -0.01 | |||
| Opponent: GPT-4 (prompt agent) | CoT | -0.04 | |||
| Opponent: GPT-3.5-turbo-1106 (prompt agent) | CoT | -0.09 | |||
| Opponent: GPT-3.5-turbo-1106 (prompt agent) | SC-CoT | -0.13 | |||
| Opponent: GPT-3.5-turbo-1106 (prompt agent) | ToT | -0.15 |
Scores at each effort level
Models the leaderboard ran at two or more effort levels. We don’t have enough cost data here to say what the extra effort costs.
Source: Official GTBench leaderboard
| Model | Effort | Harness | Score | Cost per task | Output tokens |
|---|---|---|---|---|---|
| GPT-4 Jun 2023 | Opponent: GPT-3.5-turbo-1106 (prompt agent) | Prompt | 0.13 | ||
| GPT-4 Jun 2023 | Opponent: GPT-3.5-turbo-1106 (prompt agent) | CoT | 0.13 | ||
| GPT-3.5 Turbo Jan 2024 | Opponent: GPT-3.5-turbo-1106 (prompt agent) | CoT | 0.02 | ||
| GPT-4 Jun 2023 | Opponent: GPT-4 (prompt agent) | Prompt | 0.00 | ||
| GPT-3.5 Turbo Jan 2024 | Opponent: GPT-3.5-turbo-1106 (prompt agent) | Prompt | 0.00 | ||
| GPT-3.5 Turbo Jan 2024 | Opponent: GPT-3.5-turbo-1106 (prompt agent) | SC-CoT | -0.00 | ||
| GPT-3.5 Turbo Jan 2024 | Opponent: GPT-3.5-turbo-1106 (prompt agent) | ToT | -0.01 | ||
| Opponent: GPT-3.5-turbo-1106 (prompt agent) | Prompt | -0.01 | |||
| Opponent: GPT-4 (prompt agent) | Prompt | -0.01 | |||
| Opponent: GPT-4 (prompt agent) | CoT | -0.04 | |||
| Opponent: GPT-3.5-turbo-1106 (prompt agent) | CoT | -0.09 | |||
| Opponent: GPT-3.5-turbo-1106 (prompt agent) | SC-CoT | -0.13 | |||
| Opponent: GPT-3.5-turbo-1106 (prompt agent) | ToT | -0.15 |
Number one over time
Lab claims count from the day the card came out, independent runs from the model’s release.
Source: Official GTBench leaderboard
Scores over time
All the scores we have, one dot each, so a model tested at several settings appears more than once.
Source: Official GTBench 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
- 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 2 years ago, in January 2024, for GPT-3.5 Turbo. 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 (5 entries).
- Pass Models scored 3 models we track, plus 2 other entries on its leaderboard.
- Warning Newest score The newest came 2 years ago, in January 2024, for GPT-3.5 Turbo.
- Note Frontier models None of the top 10 has a score. The strongest tested is GPT-4 (#186).
- Note Headroom Scored as a plain number, with no fixed ceiling.
Leaderboard import 3 pass
- Pass Importer We read the board on huggingface.co: 13 rows for models we track and 2 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.