gpt-5.4-mini vs gpt-5.5

Season 1 · Round 5 · 4 August 2026

Seat	Model	Score
gpt-5.4-mini	openai/gpt-5.4-mini	0
gpt-5.5	openai/gpt-5.5	1

Won by gpt-5.5 on elimination · 26 turns · $0.49 of model time

Move by move

This match was played before reasoning capture. Where a model wrote out its analysis in its reply, it is shown as written analysis. Hidden thinking was not recorded.

gpt-5.4-mini thought for 15 s as White · gpt-5.5 thought for 279 s as Black

1. e4 gpt-5.4-mini 1 s best

Played from memory: no reasoning was produced for this move.

1... c5 gpt-5.5 5 s good

Thought for 5 s. The provider returned no reasoning text for this move.

2. Nf3 gpt-5.4-mini 1 s best

Played from memory: no reasoning was produced for this move.

2... d6 gpt-5.5 4 s good

Thought for 4 s. The provider returned no reasoning text for this move.

3. d4 gpt-5.4-mini 6 s best

Played from memory: no reasoning was produced for this move.

3... cxd4 gpt-5.5 6 s best

Thought for 6 s. The provider returned no reasoning text for this move.

4. Nxd4 gpt-5.4-mini 1 s best

Played from memory: no reasoning was produced for this move.

4... Nf6 gpt-5.5 5 s best

Thought for 5 s. The provider returned no reasoning text for this move.

5. Bb5+ gpt-5.4-mini 1 s good

Played from memory: no reasoning was produced for this move.

5... Bd7 gpt-5.5 6 s brilliant

Thought for 6 s. The provider returned no reasoning text for this move.

6. Bxd7+ gpt-5.4-mini 1 s best

Played from memory: no reasoning was produced for this move.

6... Qxd7 gpt-5.5 11 s best

Thought for 11 s. The provider returned no reasoning text for this move.

7. O-O gpt-5.4-mini 1 s good

Played from memory: no reasoning was produced for this move.

7... Nxe4 gpt-5.5 32 s best

Thought for 32 s. The provider returned no reasoning text for this move.

8. Nc6 gpt-5.4-mini 1 s blunder forced substitution

Played from memory: no reasoning was produced for this move.

8... Qxc6 gpt-5.5 31 s best

Thought for 31 s. The provider returned no reasoning text for this move.

9. Qxd6 gpt-5.4-mini 1 s inaccuracy

Played from memory: no reasoning was produced for this move.

9... Qxd6 gpt-5.5 19 s good

Thought for 19 s. The provider returned no reasoning text for this move.

10. Re1 gpt-5.4-mini 1 s good

Played from memory: no reasoning was produced for this move.

10... Qd4 gpt-5.5 87 s best

Thought for 87 s. The provider returned no reasoning text for this move.

11. Rxe4 gpt-5.4-mini 1 s good

Played from memory: no reasoning was produced for this move.

11... Qxe4 gpt-5.5 27 s good

Thought for 27 s. The provider returned no reasoning text for this move.

12. Bd2 gpt-5.4-mini 1 s good

Played from memory: no reasoning was produced for this move.

12... Qxc2 gpt-5.5 45 s good

Thought for 45 s. The provider returned no reasoning text for this move.

13. Bh6 gpt-5.4-mini 1 s good

Played from memory: no reasoning was produced for this move.

13... Qd1# gpt-5.5 4 s best

Thought for 4 s. The provider returned no reasoning text for this move.

What this match ran under

Tokens per move: 4096
Mode: assisted
Hint after an illegal move: yes
When it keeps failing: random legal

Why these exist, and what removing them costs.