Inference
EN
OpenAI

o4-mini

o4-mini is a reasoning model from OpenAI. 29 benchmarks count toward its index, in 8 categories.

  • Closed weights
  • Reasoning
  • Input: image · text · file
  • Released 2025-04
RankedEnough results for an index.
IndexAll categories, out of 100.44.7±3.0Median 51.0
CoverageIndex weight that has results.100%
SpeedOutput tokens per second.122/sMedian 67/s
First tokenSeconds to the first output token.6.4 sMedian 1.3 s
Input / 1MUS dollars per 1M input tokens.$1.10Median $0.60
Output / 1MUS dollars per 1M output tokens.$4.40Median $2.55
ContextMaximum tokens in one request.200KMedian 262K
EloLMArena rating and rank.1353±4 #209Median 1422

Details

From OpenRouter
API IDThe model ID on OpenRouter.
openai/o4-mini
Max outputMaximum output tokens per request.
100K
Knowledge cutoffLast date of training data.
30 Jun 2024
ToolsTool calls on OpenRouter.
Yes
JSON outputOutput follows a given JSON schema.
Yes
Cache read / 1MUS dollars per 1M cached tokens.
$0.275
WeightsLink to the published weights.
—
RetiresDate OpenRouter removes the model.
—

Capabilities

Index per category, out of 100.
Out of 100
AgenticMulti-step tasks with tools.1 counted38.5—
CodingCode writing and repair.4 counted45.2162 / 209
ReasoningLogic problems and puzzles.6 counted44.9144 / 227
MultimodalTasks with images and text.1 counted52.0—
KnowledgeFacts and expert knowledge.7 counted40.8191 / 268
MultilingualTasks in many languages.1 counted35.5—
InstructionTasks with strict prompt rules.1 counted53.2—
MathMath problems.8 counted50.585 / 168

Results

29 counted
BenchmarkThe benchmark name.ResultThe score from the source.PlacePlace among results on this benchmark.
MATH level 5Math97.8% ±0.32 / 88
Evaluating Large Language Models Trained on CodeCoding97.3%1 / 22
Multilingual Grade School MathMultilingual93.7%1 / 13
MATH 500Math94.2% ±1.010 / 59
Massive Multitask Language UnderstandingKnowledge90.0%6 / 34
MATH-500 Problem SetMath97.5%6 / 27
Artificial Analysis IFBenchInstruction68.7%51 / 146
SWE-bench Verified · SWE-bench/experimentsCoding64.6%23 / 65
FrontierMath-2025-02-28-PrivateMath24.8% ±2.525 / 69
LiveCodeBenchCoding82.2% ±1.053 / 142
OTIS Mock AIME 2024-2025Math81.7% ±4.783 / 200
FrontierMath-Tier-4-2025-07-01-PrivateMath6.3% ±3.523 / 55
GPQA diamond · Epoch AIKnowledge79.6% ±2.491 / 217
AIMEMath83.7% ±1.241 / 95
Artificial Analysis MATH-500Math98.9%4 / 9
Artificial Analysis LiveCodeBenchCoding85.9%6 / 12
MMMU ProMultimodal79.7% ±1.046 / 92
Artificial Analysis GPQA DiamondKnowledge78.4%103 / 199
Graduate-Level Google-Proof Q&AKnowledge77.6%73 / 141
SWE-bench Verified · swebench.comCoding45.0%24 / 46
ARC-AGI-1 (semi-private)Reasoning58.7%69 / 126
Artificial Analysis Humanity's Last ExamKnowledge16.5%122 / 222
ARC-AGI-2 (semi-private)Reasoning6.1%75 / 131
τ²-Bench Tool-Agent-User EvaluationAgentic55.6%89 / 149
GPQA Diamond · Vals AIKnowledge74.5% ±2.283 / 135
Critical Physics TasksReasoning0.6%132 / 214
MMLU ProKnowledge80.6% ±0.486 / 135
Measuring Short-Form Factuality in Large Language ModelsKnowledge20.2%9 / 14
Artificial Analysis Omniscience AccuracyKnowledge24.8%106 / 164
Artificial Analysis Long Context ReasoningReasoning61.0%141 / 217
Artificial Analysis MMMU-ProMultimodal69.2%77 / 113
FrontierMath-Tiers-1-3-v2-PrivateMath36.1% ±2.962 / 90
IOI v1Coding4.8% ±2.445 / 61
τ²-bench TelecomAgentic50.2%15 / 18
Chess PuzzlesReasoning13.0% ±3.476 / 90
FrontierMath-Tier-4-v2-PrivateMath4.9% ±3.462 / 72
SimpleQA VerifiedKnowledge18.8% ±1.280 / 91
τ²-bench AirlineAgentic52.1%20 / 22
Mystery Game PuzzlesReasoning5.0% ±2.275 / 80
τ²-bench RetailAgentic68.3%22 / 23
FrontierMath-Tier-4-2025-07-01-PublicMath0.0%3 / 36
MedQAHealthcare96.0% ±0.28 / 95
MGSMMultilingual93.4% ±1.59 / 74
TaxEval v2Finance74.8% ±0.929 / 143
FrontierMath-2025-02-28-PublicMath30.0% ±15.322 / 64
SWE-bench Multimodal · swebench.comMultimodal33.9%3 / 8
SWE-bench Multimodal · SWE-bench/experimentsMultimodal33.8%3 / 7
MortgageTaxFinance64.8% ±0.946 / 97
ARC-AGI-2 (public eval)Reasoning7.5%62 / 122
ARC-AGI-1 (public eval)Reasoning68.0%60 / 116
CorpFin v2Finance59.0% ±1.079 / 131
LegalBenchLegal79.2% ±0.597 / 149
MedCodeHealthcare33.8% ±2.083 / 106
MedScribeHealthcare69.1% ±2.097 / 109
Artificial Analysis AIME 2025Math90.7%—
Discrete Reasoning Over ParagraphsReasoning77.7%—

Sources

BenchLMCC BY-NC 4.0OpenRouterTerms restrict reuseEpoch AICC BYVals AINo licence statedSierra ResearchMITSWE-bench teamNo licence statedARC Prize FoundationTerms restrict reuse

History

5 releases
Index45.5 → 44.7
23 Oct 202612 Nov 2026
Output / 1M$4.40 → $4.40
23 Oct 202612 Nov 2026

Questions

How much does o4-mini cost?
$1.10 per 1M input tokens, $4.40 per 1M output tokens.
How good is o4-mini?
Its index is 44.7 out of 100. That is place 179 of 283 ranked models.
What is o4-mini best at?
Math: 50.5 out of 100, place 85 of 168.
How fast is o4-mini?
122 output tokens per second, 6.4 s to the first token.
What is the context window of o4-mini?
200K tokens.
Is o4-mini open source?
No. Its weights are not published.
When was o4-mini released?
April 2025.

Badge

o4-mini on Inference360: place 179 of 283, index 44.7
HTML
<a href="https://inference360.ai/models/o4-mini"><img src="https://inference360.ai/badges/o4-mini.svg" alt="o4-mini on Inference360: place 179 of 283, index 44.7" height="20"></a>
Markdown
[![o4-mini on Inference360: place 179 of 283, index 44.7](https://inference360.ai/badges/o4-mini.svg)](https://inference360.ai/models/o4-mini)

Compare with

DeepSeek-R1-052844.6o144.5Qwen3 Max45.0Grok 4 Fast (Reasoning)45.3GPT-OSS 120B43.4

Lower price

dots3-note Preview63.4 · FreeApodex 1.1 Mini55.7 · FreeNemotron 3 Ultra50.8 · Free

More from OpenAI

All models
o351.8o4-mini (high)UnrankedGPT-4.138.0GPT-4.1 mini34.5GPT-4.1 nano26.3