What are o3-mini's benchmark scores?

As of 25 Oct 2026, o3-mini scores 49.9 ±4.7 on the BenchLeader Index, strongest in coding (57) and weakest in agents & tools (40). That rests on 7 independent benchmark results, and every score on this page links to its publisher.

How does o3-mini rank?

o3-mini is #385 of 785 ranked configurations on 25 Oct 2026. 323 other configurations score within its margin, so the exact position matters less than the band. The leader is Claude Opus 5.5 (max effort) at 72.6.

How much does o3-mini cost per 1M tokens?

As of 25 Oct 2026, o3-mini costs $1.10 per 1M input tokens and $4.40 per 1M output tokens, $1.93 blended (three input tokens per output token).

When was o3-mini released?

OpenAI released o3-mini on 31 Jan 2025. BenchLeader has results for it at 4 reasoning-effort settings.

OpenAIReasoning modelCompare with another model

o3-mini

Reasoning effort

o3-mini is an OpenAI proprietary reasoning model, released 31 Jan 2025. Its best configuration ranks #385 of 785 on the BenchLeader Index at 49.9 ±4.7, in the upper half. The ± is the point: 323 other configurations score within that range, so they and this one cannot be told apart on quality alone — price and speed are what separate them. It scores highest in coding (57) and lowest in agents & tools (40). At $1.93 per million tokens blended it is pricier than most ranked models. Output speed of 187 tokens per second puts it in the fastest quarter, with a first answer in 7.3 s. It has been measured at 4 reasoning-effort settings; this summary describes the best-scoring one, and the tabs above switch between them. Last measured 22 Oct 2026.

Blended price: $1.93/M; $1.10 in · $4.40 out
Output speed: 187 tok/s; measured by Artificial Analysis
First answer: 7.30 s
Context: 200k
Full answer: 9.98 s; median, reasoning included
Cost per run: –
Released: 31 Jan 2025

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

Overall index50
Reasoning52
Coding57
Agents & tools40
Maths55
Human preference51
Composite43

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. “Not stated” collects results from publishers that did not say which setting they used; for a reasoning model that is usually its thinking mode, but we do not assume it. Pick a setting here or at the top of the page to see its price, speed and category scores.

Effort	Index	Rank	Speed	First answer	Chat reply cost	Categories
low	40.4	#658	187 tok/s	7.30 s	$0.0018	Coding 36 · Maths 33 · Reasoning 38
medium	45.2	#522	187 tok/s	7.30 s	$0.0018	Agents & tools 46 · Long context 41 · Maths 46 · Reasoning 41
high	47.2	#461	225 tok/s	16 s	$0.0018	Agents & tools 35 · Coding 47 · Composite 42 · Human preference 53 · Instruction following 66 · Knowledge 41 · Long context 46 · Maths 51 · Reasoning 41
not statedbest	49.9	#385	187 tok/s	7.30 s	$0.0018	Agents & tools 40 · Coding 57 · Composite 43 · Human preference 51 · Maths 55 · Reasoning 52

How its index has moved

8 Sept 2026 to 21 Oct 2026

not stated
high
medium
low

The index is recomputed from scratch every day, so a line moves when a new benchmark result lands, when a publisher revises a score, or when the models it is normalised against change. Early movement usually means the score is still settling.

Benchmark results

See all LLM benchmarks. 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	not stated	Source
GPQA Diamond	68.2%#174	74.3%#153	77.0%#139	–	Epoch AI Benchmarking Hub
SimpleBench	–	–	22.8%#77	–	SimpleBench
Arena Hard Prompts	–	–	1401#172	1371#196	Arena (LMArena)
GPQA Diamond (AA)not in index	–	–	77.3%#222	74.8%#254	Artificial Analysis
Humanity's Last Exam (AA)not in index	–	–	12.0%#283	7.9%#356	Artificial Analysis
GPQA Diamond (Vals)not in index	–	–	75.5%#76	–	Vals AI
ARC-AGI-1	14.5%#228	22.3%#217	34.5%#199	–	ARC Prize
ARC-AGI-2	0.0%#242	2.1%#201	3.0%#193	–	ARC Prize
CritPt	–	–	0.3%#271	–	Artificial Analysis
SciPredictnot in indexeffort not stated	–	–	–	19.8%#7	Scale AI SEAL
Chess Puzzlesnot in index	6.0%#130	11.0%#118	15.0%#107	–	Epoch AI Benchmarking Hub
Mystery Game Puzzlesnot in index	–	–	7.0%#123	–	Epoch AI Benchmarking Hub
LMCAnot in index	–	–	19.0%#161	–	Epoch AI Benchmarking Hub
DTBenchnot in index	–	–	68.8%#159	–	Epoch AI Benchmarking Hub
ForecastBenchnot in indexeffort not stated	–	–	–	63.8%#38	ForecastBench

Coding

Benchmark	low	medium	high	not stated	Source
SciCode	–	–	42.8%#148	–	SciCode
WeirdML	–	–	43.7%#88	–	WeirdML
GSO-Bench	1.3%#33	–	1.3%#33	–	GSO-Bench
Arena Coding	–	–	1434#165	1416#181	Arena (LMArena)
SciCode (AA)not in index	–	–	42.8%#158	–	Artificial Analysis
LiveCodeBench	–	–	71.5%#82	–	Vals AI
Aider Polygloteffort not stated	–	–	–	60.4%#14	Aider polyglot leaderboard
SWE-bench Verified (any scaffold)not in indexeffort not stated	–	–	–	42.4%#42	SWE-bench
CadEvalnot in index	–	54.0%#5	–	–	Epoch AI Benchmarking Hub
Terminal-Bench 4.0 (AA)not in index	–	–	0.0%#152	–	Artificial Analysis
Terminal-Bench 2.1 (AA)not in index	–	–	4.5%#208	–	Artificial Analysis

Agents & tools

Benchmark	medium	high	not stated	Source
Cybench	22.5%#10	–	–	Cybench
Terminal-Bench Hard	–	6.1%#266	6.8%#249	Artificial Analysis
τ²-Bench Telecom (AA)not in index	–	31.3%#253	28.6%#266	Artificial Analysis
τ³-Banking (AA)	–	5.2%#184	–	Artificial Analysis
AutomationBenchnot in index	–	0.5%#205	–	Artificial Analysis
GDP.pdf (AA)not in index	–	2.0%#190	–	Artificial Analysis
BrowseCompnot in indexeffort not stated	–	–	1.5%#12	OpenAI

Maths

Benchmark	low	medium	high	not stated	Source
FrontierMath Tiers 1–3	3.9%#113	10.5%#104	18.6%#94	–	Epoch AI Benchmarking Hub
FrontierMath Tier 4	–	–	0.0%#72	–	Epoch AI Benchmarking Hub
OTIS Mock AIME	44.4%#193	63.9%#156	76.9%#125	–	Epoch AI Benchmarking Hub
MATH Level 5	–	95.2%#14	96.5%#10	–	Epoch AI Benchmarking Hub
AIME (Vals)	–	–	86.5%#32	–	Vals AI
MGSM	–	–	91.3%#26	–	Vals AI
MATH 500not in index	–	–	91.8%#19	–	Vals AI
Arena Maths	–	–	1405#142	1382#170	Arena (LMArena)

Knowledge

Benchmark	high	Source
SimpleQA Verified	15.3%#79	Epoch AI Benchmarking Hub
AA-Omniscience	-42.6#336	Artificial Analysis
MMLU-Pro	78.7%#96	Vals AI
LegalBench	71.5%#115	Vals AI
CorpFin	45.3%#110	Vals AI
TaxEval	69.4%#94	Vals AI
MedQA	94.8%#14	Vals AI
AA-Omniscience: accuracynot in index	21.3%#285	Artificial Analysis
AA-Omniscience: non-hallucinationnot in index	18.9%#290	Artificial Analysis

Instruction following

Benchmark	high	not stated	Source	Trend
IFBench	67.1%#89	–	Artificial Analysis
Arena Instruction Followingnot in index	1366#176	1343#194	Arena (LMArena)

Human preference

Benchmark	medium	high	not stated	Source
Arena Text	–	1364#189	1348#203	Arena (LMArena)
Arena Creative Writingnot in index	–	1311#203	1301#217	Arena (LMArena)
Arena Multi-turnnot in index	–	1345#199	1340#207	Arena (LMArena)
Arena Longer Queriesnot in index	–	1373#184	1359#196	Arena (LMArena)
Lech Mazur Creative Writingnot in index	61.5%#41	61.7%#40	–	Lech Mazur

Long context

Benchmark	medium	high	Source
Fiction.LiveBench 120k	43.8%#24	–	Fiction.live
AA-LCR	–	43.0%#347	Artificial Analysis
MLCRnot in index	–	7.2%#135	Artificial Analysis

Composite

Benchmark	high	not stated	Source	Trend
Epoch Capabilities Indexnot in indexeffort not stated	–	140.2#110	Epoch AI Benchmarking Hub
AA Intelligence Index v4.3.2	11.0#368	12.5#330	Artificial Analysis

Safety & honesty

Benchmark	low	medium	high	not stated	Source
FORTRESSnot in indexeffort not stated	–	–	–	30.1%#24	Scale AI SEAL
MASKnot in index	49.7%#50	48.9%#55	46.8%#59	–	Scale AI SEAL
PropensityBenchnot in indexeffort not stated	–	–	–	33.2%#10	Scale AI SEAL

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 $0.550 per 1M. Reasoning tokens are not modelled.

Workload	Tokens in / out	Cost	With caching	Time
Chat reply	400 / 300	$0.0018	$0.0016	8.9 s
Summarize a 30-page report	12,000 / 600	$0.016	$0.011	10.5 s
Code edit	6,000 / 1,500	$0.013	$0.011	15.3 s
Agentic coding session	60,000 / 4,000	$0.084	$0.059	28.7 s
Structured extraction	2,000 / 200	$0.0031	$0.0023	8.4 s

New this month

Questions people ask

What are o3-mini's benchmark scores?: As of 25 Oct 2026, o3-mini scores 49.9 ±4.7 on the BenchLeader Index, strongest in coding (57) and weakest in agents & tools (40). That rests on 7 independent benchmark results, and every score on this page links to its publisher.
How does o3-mini rank?: o3-mini is #385 of 785 ranked configurations on 25 Oct 2026. 323 other configurations score within its margin, so the exact position matters less than the band. The leader is Claude Opus 5.5 (max effort) at 72.6.
How much does o3-mini cost per 1M tokens?: As of 25 Oct 2026, o3-mini costs $1.10 per 1M input tokens and $4.40 per 1M output tokens, $1.93 blended (three input tokens per output token).
When was o3-mini released?: OpenAI released o3-mini on 31 Jan 2025. BenchLeader has results for it at 4 reasoning-effort settings.

Data as of 25 Oct 2026. Compare these configurations.

Cite as: BenchLeader, “o3-mini: benchmarks, pricing, speed and rank”, data as of 25 Oct 2026, https://www.benchleader.com/models/o3-mini.