Solar System
loads medium
Tech. quality98
UX92
Time5 min 46 s
Cost$0.41
Compare →
GPT-5.4 mini
Loads without errors4/5#12 of 26
Tech. quality95/100#9 of 26
UX84/100#17 of 26
Total cost$2.16
Average time6 min 04 s#19 of 26
Total tokens482,419
API price: $0.75 / $4.5 per 1M input / output tokens · as of 17/10/2026 source ↗
What is GPT-5.4 mini good for?
OpenAI's small, fast model – by the maker's account its strongest mini model yet for coding, computer use and subagents.
Description based on the maker · source ↗
- Maker
- OpenAI
- Released
- 17/03/2026
- Model weights
- closed, the model is available through the API only
- Reasoning levels
- low medium high
How did GPT-5.4 mini do in the AI Lab tests?
- 4 of 5 outputs load without errors.
- Its best technical result was on the “Monitor test” prompt (100/100).
- Did not work (original output): GTA-style game.
- It implemented the chess rules flawlessly – passing the objective perft test on all 6 positions.
- Average time per prompt 6 min 04 s#19 of 26
- Average cost per prompt $0.43#23 of 26
- Technical quality 95/100#9 of 26
- Automatic UX rating 84/100#17 of 26
Outputs
All outputs of GPT-5.4 mini on the 5 prompts every other model got as well.
GTA-style game
broken medium fixed version →
Tech. quality80
UX100
Time7 min 09 s
Cost$0.49
Compare →
Monitor test
loads medium
Tech. quality100
UX87
Time4 min 24 s
Cost$0.31
Compare →
New York & Central Park map
loads medium
Tech. quality100
UX71
Time7 min 15 s
Cost$0.53
Compare →
Chess with an engine
loads medium
Tech. quality98
UX70
Time5 min 47 s
Cost$0.42
Compare →
Cost = tokens priced with the API price list valid on the date shown next to the cost (prices change). ≈ = estimate (only the total token count is known), no ≈ = exact from input / output tokens. “Loads” = the page opens without errors; technical quality also checks behaviour during use – judge the quality in the viewer.