GPT-5.4 mini: benchmarks, price and rank
GPT-5.4 mini from OpenAI is 84th of 123 rated AI models on Spezia's leaderboard. It costs $0.75 in, $4.50 out per million tokens.
- 84th of 123 rated models
- Rating 943
- Rated on 14 benchmarks
- Closed
- Released Mar 17, 2026
- 400,000 tokens of context
- Knows the world to August 2025
Where it is strongest
Not in the top tenth on any benchmark in use.
Where it falls behind
- 60th of 66 on FrontierMath Tier 4
- 29th of 33 on MCP Atlas
- 66th of 78 on SimpleQA Verified
Price: $0.75 in, $4.50 out per million tokens.
Its headline results
The benchmarks that count most in the rating, with its place among the models measured the same way, and where each number came from.
- MCP Atlas57.7%
Multi-step tasks where the model has to pick and combine many real tools, connected through MCP, the standard way agents plug into apps and data.
29th of 33. The maker's own number, from OpenAI, via LLM Stats.
- SWE-Bench Pro54.4%
Long changes to real codebases in several languages, often across many files, including code from private and copyleft projects that models are unlikely to have seen.
36th of 50. The maker's own number, from OpenAI, via LLM Stats.
- Humanity's Last Exam28.2%
About 2,500 questions written by experts in more than a hundred fields to be as hard as they could make them, many needing a diagram or image.
43rd of 74. The maker's own number, from OpenAI, via LLM Stats.
- OSWorld-Verified72.1%
Ordinary tasks on a real desktop computer using normal apps: a browser, a spreadsheet, an email client, a photo editor, the file manager.
16th of 25. The maker's own number, from OpenAI, via LLM Stats.
- Toolathlon42.9%
Long tasks across many real apps and their tools, such as email, calendars, spreadsheets, code repositories and online stores.
28th of 37. The maker's own number, from OpenAI, via LLM Stats.
- SimpleQA Verified29.4%
Short factual questions with one checked answer, asked without web search.
66th of 78. Independent result from Epoch AI.
- FrontierMath51.2%
Hundreds of new, unpublished math problems written by mathematicians, from hard undergraduate level up to research, with answers a computer can check.
54th of 76. Independent result from Epoch AI.
- MMMU-Pro76.6%
College exam questions that need a picture to answer, such as charts, diagrams, maps and chemical structures, with more answer options than the original MMMU and some questions shown only as a screenshot.
27th of 51. The maker's own number, from OpenAI, via LLM Stats.
- MRCR v2 (8 needles)33.6%
A very long conversation full of similar requests, where the model has to find and repeat particular earlier answers, eight of them at once.
15th of 24. The maker's own number, from OpenAI, via LLM Stats.
- LMArena text1,412
People put a question to two anonymous models and vote for the better answer.
76th of 107. Independent result from LMArena.
8 more results in its full profile.
Every result with its source, what it costs per task, what jumped with its release, side by side with any other model, and answers to your questions about when to use it.
Models ranked near it
How the rating works, and every benchmark it uses: the model leaderboard.