This model is 16 months old. The same company has since released something newer: GPT-5.6-Cyber (2026-08-10). The details below are still accurate for this model — it just is not what we would recommend today.
o4-mini
OpenAI🇺🇸 United States
Fast, affordable reasoning model. Strong on STEM tasks.
Context window200K tokens
Input / 1M tokens$1.1
Output / 1M tokens$4.4
Version History
o4-mini-2025-04-16major
o4-mini brings strong reasoning at 8x lower cost than o3. API tier access.
Benchmark Scores
Full leaderboard →93.4%
AIME 2024
93.4%
AIME 2025
90.2%
DocVQA
81.4%
GPQA
18.6%
Hallucination (Summarization)
80.5%
Hallucination Rate
99.4%
HumanEval
81.8%
LiveCodeBench
99.5%
MATH
90.5%
MMLU
83.8%
MMLU-Pro
82.9%
MMMU
115.0 tokens_per_sec
Speed (tok/s)
49.5%
SWE-bench Verified