o4-mini (high): benchmark scores, price and ranking

o4-mini (high) from OpenAI ranks #485 of 743 LLMs on the LLMs Tiger aggregate benchmark score (46.8), based on 17 benchmark results captured up to 2026-10-13.

Organisation

OpenAI

Type

Fast / low-cost (flash) model

Open weights

No

Released

2025-04-16 (source: curated)

Cheapest API price

$1.00 in / $4.00 out per million tokens (via replicate, observed 2026-10-13T19:47Z)

Aggregate rank

#485 of 743 (score 46.8)

Relative to other models, o4-mini (high) is strongest on EnigmaEval (#4 of 24) and weakest on Mystery Games (#56 of 59).

Across the 17 benchmarks where at least ten models are scored, o4-mini (high) is in the top 10% on 0 and in the top half on 8.

Its cheapest listed input price, $1.00 per million tokens, is lower than that of 34% of the 427 models with a listed price.

o4-mini (high) benchmark results

About o4-mini (high)

o4-mini (high) is a smaller, lower-cost model from OpenAI. It was released on 2025-04-16, according to the developer's announcement. The cheapest API listing we track charges $1.00 per million input tokens and $4.00 per million output tokens, through replicate. We hold 17 benchmark results for it across 4 areas, and none is in the top tenth of its board.

Where o4-mini (high) stands

On our aggregate score, o4-mini (high) is #485 of 743 ranked models, in the lower half. Its score is 46.8; the scale is centred on 50, so it sits 3.2 points below the typical model. That beats 35% of the ranked field. Among the 88 ranked OpenAI models it is #67. The next one up is GPT-4 (1106) at 46.9. The next one down is GPT-4.5 Preview (Feb 2025) at 46.6.

o4-mini (high), benchmark by benchmark

EnigmaEval puts o4-mini (high) at 9.2%, #4 of 24. That beats 83% of the models tested, but 9.0 points short of Gemini 3 Pro Preview (18.2%). The results either side are o3 (high) (11.9%) and Claude Opus 4.6 (max) (7.6%). It covers multi-step puzzle hunts. The board median is 2.9%, so o4-mini (high) is 6.3 points above it. Source: Epoch AI Benchmarking Hub, captured 2026-10-13.

On WeirdML, o4-mini (high) reaches 52.6%. It ranks #26 of 109, ahead of 76% of the field, and 40.7 points behind GPT-6 Astra (max) at 93.3%. The results either side are GPT-5 mini (high) (52.7%) and o3 (high) (52.4%). It covers unusual machine-learning tasks. The board median is 41.2%, so o4-mini (high) is 11.4 points above it. Source: Epoch AI Benchmarking Hub, captured 2026-10-13.

o4-mini (high) manages 26.0% on Chess Puzzles, good for #31 of 122. GPT-6 Astra (max) leads at 72.0%, 46.0 points further up, and 75% of models sit below. The results either side are Kimi K2.6 (26.0%) and Qwen3.6 35B-A3B (26.0%). It covers chess tactics puzzles. The board median is 12.5%, so o4-mini (high) is 13.5 points above it. Source: Epoch AI Benchmarking Hub, captured 2026-10-13.

OTIS Mock AIME puts o4-mini (high) at 81.7%, #64 of 189. That beats 66% of the models tested, but 18.3 points short of Claude Fable 5.1 (max) (100.0%). The results either side are Gemma 4 26B A4B (82.2%) and GPT-5 nano (high) (81.1%). It covers competition maths (AIME 2024-25 style). The board median is 57.8%, so o4-mini (high) is 23.9 points above it. Source: Epoch AI Benchmarking Hub, captured 2026-10-13.

On Humanity's Last Exam, o4-mini (high) reaches 18.1%. It ranks #10 of 27, ahead of 63% of the field, and 22.5 points behind Muse Spark at 40.6%. The results either side are Gemini 2.5 Pro Exp (Mar 2025) (18.2%) and Gemini 2.5 Flash Preview (Apr 2025) (12.1%). It covers expert-level questions across domains. The board median is 10.7%, so o4-mini (high) is 7.4 points above it. Source: Epoch AI Benchmarking Hub, captured 2026-10-13.

o4-mini (high) manages 79.6% on GPQA Diamond, good for #107 of 275. GPT-6 Astra (max) leads at 95.8%, 16.2 points further up, and 61% of models sit below. The results either side are Claude 3.7 Sonnet (64k thinking) (79.7%) and NVIDIA-Nemotron-3-Super-120B-A12B-BF16 (79.2%). It covers expert-level science questions. The board median is 71.7%, so o4-mini (high) is 7.9 points above it. Source: Epoch AI Benchmarking Hub, captured 2026-10-13.

o4-mini (high) lands below the middle on GSO-Bench: 3.6%, #15 of 19, with 74% of the field scoring higher. GPT-5.5 (xhigh) leads at 40.2%; the closest result above is claude-3-7-sonnet-20250219 at 3.8%. It covers software performance optimisation. The board median is 4.9%, so o4-mini (high) is 1.3 points below it. Source: Epoch AI Benchmarking Hub, captured 2026-10-13.

SimpleQA Verified is a weak spot. o4-mini (high) scores 18.8%, #59 of 69, behind 84% of the models measured. The leader, gemini-4-argon-high, reaches 76.0%. It covers short factual questions. The board median is 41.0%, so o4-mini (high) is 22.2 points below it. Source: Epoch AI Benchmarking Hub, captured 2026-10-13.

FrontierMath T4 is a weak spot. o4-mini (high) scores 4.9%, #45 of 50, behind 88% of the models measured. The leader, GPT-6.1 Sol (max), reaches 100.0%. It covers research-level maths, tier 4. The board median is 30.5%, so o4-mini (high) is 25.6 points below it. Source: Epoch AI Benchmarking Hub, captured 2026-10-13.

Mystery Games is a weak spot. o4-mini (high) scores 5.0%, #56 of 59, behind 93% of the models measured. The leader, GPT-6 Astra (max), reaches 84.0%. It covers deductive reasoning puzzles. The board median is 23.0%, so o4-mini (high) is 18.0 points below it. Source: Epoch AI Benchmarking Hub, captured 2026-10-13.

The full table above also lists the other 7 results, including boards with fewer than ten models, where a rank says little.

Strengths and weaknesses

Software engineering: uneven, ahead of 40% of the field on average across 2 benchmarks. Its best board here is ALE-Bench (#32 of 79) and its weakest is GSO-Bench (#15 of 19). Reasoning, maths and science: uneven, ahead of 48% of the field on average across 9 benchmarks. Its best board here is EnigmaEval (#4 of 24) and its weakest is FrontierMath T4 (#45 of 50). General knowledge and chat: uneven, ahead of 43% of the field on average across 5 benchmarks. Its best board here is WeirdML (#26 of 109) and its weakest is Mystery Games (#56 of 59).

Price and value

The median input price across the 427 priced models is $0.500 per million tokens, so o4-mini (high) costs more than the typical listing. Models that score within 1.5 points or better and cost less: Qwen2.5-Coder-7B-Instruct ($0.010, 49.5), DeepSeek V4.1 Flash (high) ($0.010, 48.9), Qwen2.5-Coder-3B-Instruct ($0.010, 46.5). Similar scores cost more on mimo-v2-pro ($1.10) and o3-mini-2025-01-31 (low) ($1.10).

Release history

The OpenAI model released before it was o3 (high) on 2025-04-16, which scores 49.8. The next one was gpt-oss-120b (high) on 2025-08-05, at 47.3.

Reading these numbers

The 17 scores on this page come from 2 sources: Epoch AI Benchmarking Hub (16), ARC Prize official leaderboard (1). 1 of them come from an official leaderboard and 16 from datasets or third-party runs. The leaderboard also lists 3 other reasoning-effort settings of this model, which redirect here. This page shows the highest.

o4-mini (high) is also listed at 3 other reasoning-effort settings; this page shows the highest-effort row, which is the one the leaderboard keeps by default.

Models ranked nearby

• #482 qwen3-30b-a3b-instruct-2507 (46.8)
• #483 Command A+ (46.8)
• #484 MiniMax-M2.5 (46.8)
• #486 codestral-2508 (46.8)
• #487 Mellum2-12B-A2.5B-Instruct (46.8)
• #488 Gemini 2.0 Pro Exp (Feb 2025) (46.7)

More from OpenAI

• GPT-6 Astra (max) (#2, 64.0)
• GPT-6.1 Sol (max) (#5, 60.6)
• GPT-5.6 Sol (max) (#6, 60.1)
• GPT-6 Sol (max) (#7, 60.0)
• GPT-5.5 (xhigh) (#11, 58.0)
• GPT-5.6 Terra (max) (#12, 57.8)

Compare it on the live leaderboard → · How the aggregate score works