GPT-5 mini

GPT-5 mini is a reasoning model from OpenAI in the GPT-5 family. 23 benchmarks count toward its score, in 7 categories.

availableShows if the model has enough results for an index.
IndexOverall score out of 100.49.8 ±3.9
CoverageShare of the index weight with results.95%
SpeedOutput tokens per second.75/s
Input / 1MUS dollars per 1M input tokens.$0.25 batch $0.125
Output / 1MUS dollars per 1M output tokens.$2 batch $1 US dollars per 1M output tokens in a batch.
ContextMaximum tokens in one request.400K
EloLMArena rating and rank.1373 (#166)

The index is a score out of 100. The ± range shows how much it can change. Batch work costs less.

26,569 votes. Elo shows what people prefer. It does not change the score.

CapabilitiesScore per category, out of 100.

Out of 100
AgenticMulti-step tasks with tools.
39.4
CodingCode writing and repair.
50.2
ReasoningLogic problems and puzzles.
49.1
MultimodalTasks with images and text.
55.8
KnowledgeFacts and expert knowledge.
47.2
MultilingualTasks in many languages.
75.5
InstructionTasks with strict rules in the prompt.
N/A
MathMath problems.
53.6

Results

23 counted
BenchmarkThe test name.CategoryThe capability that the test measures.ResultThe score from the publisher.IndexThis result as a score out of 100.RunThe settings of the run.DateDate of the result.Published byThe source of the result.
MATH level 5Math97.8%46.8high effortEpoch AI
MATH 500Math94.8%47.79 Jan 2026Vals AI
MGSMMultilingual92.6%9 Jan 2026Vals AI
AIMEMath91.5%56.416 Apr 2026Vals AI
OTIS Mock AIME 2024-2025Math86.7%60.0high effortEpoch AI
LiveCodeBenchCoding86.6%63.01 Sept 2026Vals AI
MMLU ProKnowledge82.2%50.11 Sept 2026Vals AI
GPQA DiamondKnowledge80.3%52.41 Sept 2026Vals AI
MMMU ProMultimodal78.9%55.81 Sept 2026Vals AI
GPQA diamondKnowledge75.0%47.5high effortEpoch AI
EuroEval SwedishMultilingual66.0%84.7EuroEval
SWE-Bench verifiedCoding64.7%49.4medium effortEpoch AI
SWE-benchCoding60.8%46.31 Sept 2026Vals AI
EuroEval ItalianMultilingual60.1%77.4EuroEval
EuroEval FrenchMultilingual60.1%77.3EuroEval
EuroEval DutchMultilingual60.1%77.3EuroEval
SWE-bench VerifiedCoding59.8%45.5medium effort · mini-SWE-agent26 Feb 2026SWE-bench team
SWE-bench VerifiedCoding59.8%45.5medium effort · mini-SWE-agent1 Sept 2026SWE-bench team
EuroEval PolishMultilingual57.2%73.7EuroEval
EuroEval PortugueseMultilingual55.9%72.0EuroEval
ARC-AGI-1 (semi-private)Reasoning54.3%52.9high effortARC Prize Foundation
EuroEval SpanishMultilingual54.2%70.0EuroEval
EuroEval GermanMultilingual52.5%67.9EuroEval
FrontierMath-Tiers-1-3-v2-PrivateMath46.7%57.7high effortEpoch AI
SWE-bench MultilingualMultilingual39.7%mini-SWE-agent20 Feb 2026SWE-bench team
SWE-bench MultilingualCoding39.7%mini-SWE-agent2 Sept 2026SWE-bench team
Terminal-Bench 1.0Agentic30.0%37.012 Jan 2026Vals AI
Chess PuzzlesReasoning30.0%62.3high effortEpoch AI
FrontierMath-2025-02-28-PrivateMath27.2%56.0high effortEpoch AI
Terminal-Bench 2.0Agentic27.0%41.94 Jun 2026Vals AI
SimpleQA VerifiedKnowledge21.6%41.5high effortEpoch AI
Vibe Code Bench v1.1Coding14.2%48.1OpenHands21 Sept 2026Vals AI
FrontierMath-Tier-4-v2-PrivateMath12.2%53.8high effortEpoch AI
IOI v1Coding6.8%43.89 Aug 2026Vals AI
FrontierMath-Tier-4-2025-07-01-PrivateMath6.3%50.2high effortEpoch AI
Mystery Game PuzzlesReasoning5.0%39.2high effortEpoch AI
ARC-AGI-2 (semi-private)Reasoning4.4%42.0high effortARC Prize Foundation

23 benchmarks count, from 34 of 37 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 AIEuroEval, collected directlyMIT — the leaderboard site and its CSV routes are in the licensed repositorySWE-bench team, collected directlyNo licence stated for the leaderboard. The harness repo is MITARC Prize Foundation, collected directlyNo licence stated. Their terms ask for written permission before commercial use

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