ChessBench (@chessbench) header
ChessBench (@chessbench) profile picture

A new benchmark tracking how well language models play chess. Watch the games, follow the reasoning move by move, track the leaderboard.

Joined June 2026
Thanks @kaggle, @GoogleDeepMind, and @zapier for hosting such a great event this week!
9
323
Gemini 4 Neon has been added to chessbench.ai/leaderboard, landing as the new #2 overall model! Well done @GoogleDeepMind! Somewhat unbelievably, Gemini 4 Neon at Low is on par with Claude Fable 5.2 at xhigh and GPT 6.2 Astra at xhigh...
1
40
Where will GPT 6.2 Astra land on ChessBench (by overall score)?
75%1
0%2
0%3
25%4+
4 votes • Final results
1
1
95
Where will GPT 6.2 Sol land on ChessBench (by overall score)?
33%1
33%2
0%3
33%4+
3 votes • Final results
1
1
88
ChessBench retweeted
New neurons research: introducing Neuron Communities. We discovered communities of neurons that work together inside LLMs. They’re surprisingly small, but powerful – just a few neurons work together to perform common behaviors, like arithmetic, ALL CAPS, or recalling a date. Work led by research fellow @_igorshilov:
21
106
24
882
46,322
Is there a harder task for language models, or humans, than picking random numbers?
2
1
88
Curious how well language models generate random numbers? This is going to be a tough hill to climb... benjaminbrumfield.github.io/…
50
Gemini 4 Carbon and Gemini 4 Hydrogen have been added to ChessBench, landing at second and eighth overall, respectively. Looks like the only thing keeping Carbon from the top spot is how often it times out... Well done @GoogleDeepMind !
3
6
227
In the vast majority of games it timed out on, Carbon was in roughly equal or winning positions.
1
14
Where will Gemini 4 Carbon land on ChessBench (by overall score)?
45%1
18%2
27%3
9%4+
11 votes • Final results
2
5
136
Gemini 4 Carbon has been added to ChessBench! chessbench.ai/leaderboard
1
16
Carbon is very strong but extreeeeeeemely slow. Results are going to be delayed......
2
17