Command Palette

Search for a command to run...

o4-mini (high)

OpenAI

Frontier ModelReasoningVision ModelProprietary

Context

Release Date
Apr 16, 2025
Window
200k

PricingPer 1M tokens

Input
$1.1
Output
$4.4
Blended 3:1
$1.925

Capabilities

Speed
149 t/s
Input
Output
Reasoning tokens

Latency

TTFT
38.90 ms
500 token response
42.27 s

Benchmarks

Reasoning
●●●○○
Math
●●●●●
Coding
●●●○○
MMLU Pro
83.2%
GPQA
78.4%
HLE
17.5%
SciCode
46.5%
AIME
94.0%
MATH 500
98.9%
LiveCodeBench
80.4%
HumanEval
99.0%

OpenAI o4-mini is a compact multimodal model that handles text + images, runs 200 k-token contexts, and matches GPT-4–level accuracy on math and coding while costing far less — $1.1 per M input token. It’s tuned for disciplined reasoning—refusing unsafe requests, following system instructions, and using built-in tools (web, Python, image crop) when they sharpen the answer.

For developers, o4-mini means one cheaper endpoint that can chat, write code, analyze images, and trigger tool calls without stitching services together. You get fast first-token latency, high output speed, and pricing that lets you ship larger contexts, richer agents, and production apps on a lean budget.