OpenAI

o4-mini-2025-04-16

Made by OpenAI. Released 2025-04-16. Accessibility not recorded. Cheapest standard rate we hold: $8 per million output tokens, at data sharing context.

What it costs

Tier	Kind	In	Out	Cached in	From
batch	text	$2	$8	$0.50	seller
batch	text · data sharing	$1	$4	$0.25	seller
standard	text	$4	$16	$1	seller
standard	text · data sharing	$2	$8	$0.50	seller

per million tokens, US dollars. * marks a rate from a single registry or two that disagree — the best available number rather than a confirmed one. Every other row is confirmed against a seller or a licensed index.

What it scores, by house

Benchmark	House	Score	Evaluated	Scaffold
chess puzzles	Epoch AI	0.2±0.04	2026-08-07	medium
chess puzzles	Epoch AI	0.1±0.03	2026-07-11	low
chess puzzles	Epoch AI	0.3±0.04	2025-12-08	high
frontiermath	Epoch AI	0.1±0.02	2025-11-16	low
frontiermath	Epoch AI	0.2±0.02	2025-11-13	medium
frontiermath	Epoch AI	0.2±0.03	2025-11-13	high
frontiermath tier 4	Epoch AI	0.0±0.02	2025-08-07	medium
frontiermath tier 4	Epoch AI	0.1±0.04	2025-07-01	high
frontiermath tier 4 v2	Epoch AI	0.0±0.03	2026-06-11	high
frontiermath tiers 1 3 v2	Epoch AI	0.2±0.02	2026-08-27	low
frontiermath tiers 1 3 v2	Epoch AI	0.3±0.03	2026-08-27	medium
frontiermath tiers 1 3 v2	Epoch AI	0.4±0.03	2026-06-11	high
gpqa diamond	Epoch AI	0.8±0.03	2026-08-07	medium
gpqa diamond	Epoch AI	0.8±0.03	2026-07-11	low
gpqa diamond	Epoch AI	0.8±0.02	2025-04-16	high
math level 5near its ceiling	Epoch AI	1.0±0.00	2025-04-16	high
mystery game puzzles	Epoch AI	0.1±0.02	2026-08-27	high
otis mock aime 2024 2025near its ceiling	Epoch AI	0.7±0.07	2026-08-07	medium
otis mock aime 2024 2025near its ceiling	Epoch AI	0.6±0.07	2026-07-13	low
otis mock aime 2024 2025near its ceiling	Epoch AI	0.8±0.05	2025-04-16	high
simpleqa verified	Epoch AI	0.2±0.01	2026-08-27	low
simpleqa verified	Epoch AI	0.2±0.01	2026-08-27	high
aider polyglot	collected by Epoch AI	72.0	—	high
ale bench	collected by Epoch AI	826	—	high
algotune	collected by Epoch AI	1.7	—	high
arc aginear its ceiling	collected by Epoch AI	0.4	—	medium
arc aginear its ceiling	collected by Epoch AI	0.6	—	high
arc aginear its ceiling	collected by Epoch AI	0.2	—	low
arc agi 2	collected by Epoch AI	0.1	—	high
arc agi 2	collected by Epoch AI	0.0	—	medium
arc agi 2	collected by Epoch AI	0.0	—	low
cad eval	collected by Epoch AI	0.6	—	medium
critpt	collected by Epoch AI	0.0	—	high
dtbenchnear its ceiling	collected by Epoch AI	0.8	—	high
enigma eval	collected by Epoch AI	0.1	—	medium
enigma eval	collected by Epoch AI	0.1	—	high
fictionlivebench	collected by Epoch AI	0.6	—	medium
forecastbench	collected by Epoch AI	61.8	—	unknown
gdpval	collected by Epoch AI	0.3	—	high
geobench	collected by Epoch AI	3651	—	medium
geobench	collected by Epoch AI	3717	—	high
gso	collected by Epoch AI	0.0	—	OpenHands
hle	collected by Epoch AI	0.1	—	medium
hle	collected by Epoch AI	0.2	—	high
lech mazur writing	collected by Epoch AI	7.5	—	medium
lmca	collected by Epoch AI	26.5	—	high
metr time horizons	collected by Epoch AI	76.5	—	medium
simplebench	collected by Epoch AI	0.4	—	high
vpct	collected by Epoch AI	0.6	—	medium
weirdml	collected by Epoch AI	0.5	—	high

These scores are not comparable between houses and are not added up. A score is dated by when it was evaluated, not when the model was released, and a score without its scaffold is not a measurement — which is why both are printed.

What retires, and when

deprecation — 2026-10-23
deprecation — 2026-10-23

Every rate we hold · Why there is no single best model · Everything from OpenAI