OpenAIReasoning model

o4-mini

Name: o4-mini benchmark results
Creator: BenchLeader

Best configuration ranks #188 of 610 on the BenchLeader Index at 54.9 ±4.3. Last measured 2 Sept 2026. Released 16 Apr 2025.

Blended price: $1.93/M; $1.10 in · $4.40 out
Output speed: 148 tok/s
First answer: 23 s; first token 4.24 s
Context: 200k

How it scores by categoryDashed line = average model (50). One step of 15 = one standard deviation.

Overall index55
Reasoning59
Coding55
Agents & tools58
Knowledge43
Instruction following67
Human preference57
Multimodal56
Long context57
Composite50

Reasoning-effort configurations

The same model behaves differently depending on how much it is allowed to think. Each row is one setting, scored only on the benchmarks that were run at that setting. “Default” means the publisher did not say which setting was used.

Effort	Index	Rank	Speed	First answer	Chat reply cost	Categories
low	42.4	#460	148 tok/s	23 s	$0.0018	Knowledge 33 · Maths 41 · Reasoning 42
medium	52.4	#248	148 tok/s	23 s	$0.0018	Long context 54 · Maths 48 · Multimodal 66 · Reasoning 45
high	48.0	#335	148 tok/s	23 s	$0.0018	Agents & tools 40 · Coding 49 · Instruction following 33 · Knowledge 47 · Maths 52 · Multimodal 66 · Reasoning 47
defaultbest	54.9	#188	148 tok/s	23 s	$0.0018	Agents & tools 58 · Coding 55 · Composite 50 · Human preference 57 · Instruction following 67 · Knowledge 43 · Long context 57 · Multimodal 56 · Reasoning 59

Benchmark results

One column per reasoning effort. Rank is among every configuration of every model on that benchmark. Hover a score for the run it came from.

Reasoning

Benchmark	low	medium	high	default	Source
GPQA Diamond	75.3%#137	77.8%#123	79.6%#116	–	Epoch AI Benchmarking Hub
Humanity's Last Exam	–	14.3%#22	18.1%#21	–	Scale AI / CAIS
SimpleBench	–	–	38.7%#61	–	SimpleBench
LMArena Hard Prompts	–	–	–	1406#147	LMArena
GPQA Diamond (AA)not in index	–	–	–	78.4%#205	Artificial Analysis
Humanity's Last Exam (AA)not in index	–	–	–	16.5%#183	Artificial Analysis
GPQA Diamond (Vals)not in index	–	–	74.5%#80	–	Vals AI
Kagi LLM Benchmark	–	–	–	67.6%#32	Kagi LLM Benchmark
ARC-AGI-1	21.3%#155	41.8%#129	58.7%#110	–	ARC Prize
ARC-AGI-2	1.7%#148	2.4%#140	6.1%#115	–	ARC Prize

Coding

Benchmark	low	medium	high	default	Source
WeirdML	–	–	52.6%#61	–	WeirdML
GSO-Bench	–	–	3.6%#27	–	GSO-Bench
LMArena Coding	–	–	–	1433#149	LMArena
LiveCodeBench	–	–	82.2%#53	–	Vals AI
IOI	–	–	4.8%#44	–	Vals AI
Aider Polyglot	–	–	–	72.0%#7	Aider polyglot leaderboard
SWE-bench Verified (bash only)	–	–	–	45.0%#33	SWE-bench
SWE-bench Verified (any scaffold)not in index	–	–	–	64.6%#23	SWE-bench

Agents & tools

Benchmark	low	medium	high	default	Source
GDPval	–	–	25.3%#8	–	OpenAI
Terminal-Bench Hard	–	–	–	15.2%#186	Artificial Analysis
τ²-Bench Telecom (AA)not in index	–	–	–	55.6%#179	Artificial Analysis
BFCL Overall	–	–	–	53.2%#19	Berkeley Function Calling Leaderboard

Maths

Benchmark	low	medium	high	default	Source
FrontierMath Tiers 1–3	16.1%#86	28.8%#61	36.1%#51	–	Epoch AI Benchmarking Hub
FrontierMath Tier 4	–	–	4.9%#51	–	Epoch AI Benchmarking Hub
OTIS Mock AIME	57.8%#150	73.3%#114	81.7%#97	–	Epoch AI Benchmarking Hub
MATH Level 5	–	–	97.8%#4	–	Epoch AI Benchmarking Hub
AIME (Vals)	–	–	83.7%#41	–	Vals AI
MGSM	–	–	93.4%#9	–	Vals AI
IMO 2025	–	–	14.3%#5	–	MathArena

Knowledge

Benchmark	low	medium	high	default	Source
SimpleQA Verified	19.6%#63	–	18.8%#65	–	Epoch AI Benchmarking Hub
AA-Omniscience	–	–	–	-35.7#236	Artificial Analysis
MMLU-Pro	–	–	80.6%#83	–	Vals AI
LegalBench	–	–	79.2%#89	–	Vals AI
CorpFin	–	–	59.0%#77	–	Vals AI
TaxEval	–	–	74.8%#28	–	Vals AI
MedQA	–	–	96.0%#9	–	Vals AI
PRBench Finance	–	–	–	39.2%#23	Scale AI SEAL
PRBench Legal	–	–	–	38.1%#27	Scale AI SEAL
MultiNRC	–	–	22.2%#31	–	Scale AI SEAL

Instruction following

Benchmark	low	medium	high	default	Source
IFBench	–	–	–	68.7%#78	Artificial Analysis
MultiChallenge	–	–	44.9%#25	–	Scale AI SEAL

Human preference

Benchmark	low	medium	high	default	Source
LMArena Text	–	–	–	1391#140	LMArena

Multimodal

Benchmark	low	medium	high	default	Source
LMArena Vision	–	–	–	1194#77	LMArena
MMMU-Pro	–	–	–	69.3%#121	Artificial Analysis
VISTA	–	51.7%#5	51.8%#4	–	Scale AI SEAL
MMMU (validation)	–	–	–	81.6%#4	MMMU

Long context

Benchmark	low	medium	high	default	Source
Fiction.LiveBench 120k	–	62.5%#13	–	–	Fiction.live
AA-LCR	–	–	–	61.0%#215	Artificial Analysis

Composite

Benchmark	low	medium	high	default	Source
Epoch Capabilities Indexnot in index	–	–	–	145.7#68	Epoch AI Benchmarking Hub
AA Intelligence Index	–	–	–	16.6#208	Artificial Analysis

Where it wins

Benchmarks where this configuration ranks in the top five of every configuration measured.

#4MMMU (validation)81.6% · of 46

Where to run it

Every provider serving this model through OpenRouter, with throughput and first-token latency measured on live traffic over the last 30 minutes and each provider’s own price. Purple marks the best in each column.

Provider	Speed	First token	Input $/M	Output $/M	Context	Quantisation
OpenAI	110 tok/s	4.24 s	$1.10	$4.40	200k	–

What a task costs

Estimates from list price, output speed and time to first answer for the best configuration. “With caching” assumes three-quarters of the input is served from the prompt cache at $0.275 per 1M. Reasoning tokens are not modelled.

Workload	Tokens in / out	Cost	With caching	Time
Chat reply	400 / 300	$0.0018	$0.0015	25.4 s
Summarise a 30-page report	12,000 / 600	$0.016	$0.0084	27.4 s
Code edit	6,000 / 1,500	$0.013	$0.0095	33.5 s
Agentic coding session	60,000 / 4,000	$0.084	$0.046	50.4 s
Structured extraction	2,000 / 200	$0.0031	$0.0018	24.7 s