o3

o3 is a reasoning model from OpenAI. 26 benchmarks count toward its score, in 7 categories.

availableShows if the model has enough results for an index.
IndexOverall score. 50 is the middle.53.6 ±3.7
CoverageShare of the index weight with results.95%
SpeedOutput tokens per second.77/s
Input / 1MUS dollars per 1M input tokens.$2 batch $1
Output / 1MUS dollars per 1M output tokens.$8 batch $4 US dollars per 1M output tokens in a batch.
ContextMaximum tokens in one request.200K
EloLMArena rating and rank.1410 (#130)

50 is the middle of the board. The range shows the doubt in the index. Batch work costs less.

58,579 votes. Elo shows what people prefer. It does not change the score.

CapabilitiesScore per category. 50 is the middle.

50 is the middle
AgenticMulti-step tasks with tools.
51.8
CodingCode writing and repair.
52.5
ReasoningLogic problems and puzzles.
55.3
MultimodalTasks with images and text.
55.5
KnowledgeFacts and expert knowledge.
55.2
MultilingualTasks in many languages.
N/A
InstructionTasks with strict rules in the prompt.
61.5
MathMath problems.
50.0

Results

26 counted
BenchmarkThe test name.CategoryThe capability that the test measures.ResultThe score from the publisher.IndexThis result on the index scale.RunThe settings of the run.DateDate of the result.Published byThe source of the result.
Artificial Analysis MATH-500Math99.2%Artificial Analysis
MATH level 5Math97.8%46.7high effortEpoch AI
MATH 500Math94.6%47.59 Jan 2026Vals AI
MGSMMultilingual91.7%9 Jan 2026Vals AI
MMLU ProKnowledge85.6%55.41 Sept 2026Vals AI
AIMEMath85.3%53.516 Apr 2026Vals AI
GPQA DiamondKnowledge84.1%55.91 Sept 2026Vals AI
LiveCodeBenchCoding83.9%60.61 Sept 2026Vals AI
OTIS Mock AIME 2024-2025Math83.9%58.5high effortEpoch AI
Artificial Analysis GPQA DiamondKnowledge82.7%53.3Artificial Analysis
GPQA diamondKnowledge81.8%53.8high effortEpoch AI
τ²-Bench Tool-Agent-User EvaluationAgentic80.7%56.7Victor Barres et al.
MMMU ProMultimodal80.4%58.21 Sept 2026Vals AI
Artificial Analysis Long Context ReasoningReasoning74.7%59.9Artificial Analysis
τ²-bench RetailAgentic73.9%51.8OpenAI30 Apr 2026Sierra Research
Artificial Analysis IFBenchInstruction71.4%61.5Artificial Analysis
Artificial Analysis MMMU-ProMultimodal70.1%52.8Artificial Analysis
SWE-Bench verifiedCoding62.3%47.5medium effortEpoch AI
ARC-AGI-1 (semi-private)Reasoning60.8%55.9high effortARC Prize Foundation
SWE-bench VerifiedCoding58.4%44.4mini-SWE-agent26 Feb 2026SWE-bench team
SWE-bench VerifiedCoding58.4%44.4mini-SWE-agent1 Sept 2026SWE-bench team
τ²-bench AirlineAgentic52.0%36.1OpenAI2 Mar 2026Sierra Research
SimpleQA VerifiedKnowledge49.4%67.3high effortEpoch AI
Artificial Analysis Omniscience AccuracyKnowledge38.6%61.6Artificial Analysis
SWE-bench MultimodalMultimodal36.0%GUIRepair17 Nov 2025SWE-bench team
SWE-bench MultimodalMultimodal36.0%GUIRepair17 Nov 2025SWE-bench team
Chess PuzzlesReasoning34.0%67.5high effortEpoch AI
FrontierMath-Tiers-1-3-v2-PrivateMath33.3%50.2high effortEpoch AI
Mystery Game PuzzlesReasoning29.0%64.8high effortEpoch AI
Artificial Analysis Intelligence IndexKnowledge20.2%47.7Artificial Analysis
Artificial Analysis Humanity's Last ExamKnowledge20.1%46.4Artificial Analysis
FrontierMath-2025-02-28-PrivateMath18.7%48.1high effortEpoch AI
ARC-AGI-2 (semi-private)Reasoning6.5%43.0high effortARC Prize Foundation
FrontierMath-Tier-4-2025-07-01-PrivateMath2.1%45.7high effortEpoch AI
Critical Physics TasksReasoning1.1%40.7Artificial Analysis

26 benchmarks count, from 31 of 35 results. A grey row does not count. Too few models took that benchmark.

Sources

BenchLM benchmark aggregationUsed with attribution; per-benchmark results credited to their original authorsOpenRouter, collected directlyNo licence statedEpoch AI, collected directlyCC BY — free to use and redistribute with attributionVals AI, collected directlyNo licence stated. Read from the public leaderboard and credited to Vals AISierra Research, collected directlyMIT — results are in the licensed repositoryARC Prize Foundation, collected directlyNo licence stated. Their terms ask for written permission before commercial useSWE-bench team, collected directlyNo licence stated for the leaderboard. The harness repo is MIT

Same level, lower price

Qwen3.8-Flash-Next66.4 · Freedots3-note Preview65.5 · FreeApodex 1.1 Mini59.1 · Free

More from OpenAI

GPT-6 Astra83.5GPT-5.6 Sol76.7GPT-5.6 Terra73.4GPT-6 Sol76.9GPT-5.572.3