Inference
EN
OpenAI

GPT-5 mini

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

  • Closed weights
  • Reasoning
  • Input: text · image · file
  • Released 2025-08
  • Retires 11 Dec 2026
RankedEnough results for an index.
IndexAll categories, out of 100.48.4±3.9Median 51.0
CoverageIndex weight that has results.95%
SpeedOutput tokens per second.78/sMedian 67/s
First tokenSeconds to the first output token.2.6 sMedian 1.3 s
Input / 1MUS dollars per 1M input tokens.$0.25batch $0.125Median $0.60
Output / 1MUS dollars per 1M output tokens.$2.00batch $1.00 Batch price per 1M output tokens.Median $2.55
ContextMaximum tokens in one request.400KMedian 262K
EloLMArena rating and rank.1373±5 #185Median 1422

Details

From OpenRouter
API IDThe model ID on OpenRouter.
openai/gpt-5-mini
Max outputMaximum output tokens per request.
128K
Knowledge cutoffLast date of training data.
31 May 2024
ToolsTool calls on OpenRouter.
Yes
JSON outputOutput follows a given JSON schema.
Yes
Cache read / 1MUS dollars per 1M cached tokens.
$0.025
WeightsLink to the published weights.
—
RetiresDate OpenRouter removes the model.
11 Dec 2026

Where to run

APIs that serve this model.
5 providers
ProviderTypeMaker, cloud, host or router.InputOutputContext
DatabricksCloud$0.25$2.00400K
DigitalOceanCloud$0.25$2.00400K
OpenRouterRouter$0.25$2.00400K
SAP AI CoreCloud$0.25$2.00400K
Snowflake CortexCloud$0.25$2.00272K

US dollars per 1M tokens, list price. From models.dev (opens in a new tab) (MIT).

Capabilities

Index per category, out of 100.
Out of 100
AgenticMulti-step tasks with tools.2 counted41.4174 / 218
CodingCode writing and repair.4 counted49.4142 / 209
ReasoningLogic problems and puzzles.4 counted47.0136 / 227
MultimodalTasks with images and text.1 counted57.9—
KnowledgeFacts and expert knowledge.3 counted46.1163 / 268
MultilingualTasks in many languages.1 counted51.8—
InstructionTasks with strict prompt rules.None counted——
MathMath problems.8 counted52.277 / 168

Results

23 counted
BenchmarkThe benchmark name.ResultThe score from the source.PlacePlace among results on this benchmark.
MATH level 5Math97.8% ±0.32 / 88
EuroEval SwedishMultilingual66.0% ±0.717 / 195
MATH 500Math94.8% ±1.06 / 59
EuroEval PolishMultilingual57.2% ±0.329 / 209
EuroEval DutchMultilingual60.1% ±0.425 / 175
EuroEval GermanMultilingual52.5% ±0.628 / 191
LiveCodeBenchCoding86.6% ±1.021 / 142
EuroEval SpanishMultilingual54.2% ±0.628 / 176
EuroEval ItalianMultilingual60.1% ±0.328 / 174
EuroEval FrenchMultilingual60.1% ±0.438 / 178
EuroEval PortugueseMultilingual55.9% ±0.445 / 182
AIMEMath91.5% ±0.824 / 95
SWE-bench Verified · swebench.comCoding59.8%13 / 46
FrontierMath-2025-02-28-PrivateMath27.2% ±2.622 / 69
OTIS Mock AIME 2024-2025Math86.7% ±4.369 / 200
FrontierMath-Tier-4-2025-07-01-PrivateMath6.3% ±3.523 / 55
SWE-bench Verified · SWE-bench/experimentsCoding59.8%31 / 65
GPQA Diamond · Vals AIKnowledge80.3% ±2.066 / 135
GPQA diamond · Epoch AIKnowledge75.0% ±2.3107 / 217
MMMU ProMultimodal78.9% ±1.050 / 92
ARC-AGI-1 (semi-private)Reasoning54.3%72 / 126
MMLU ProKnowledge82.2% ±0.478 / 135
Chess PuzzlesReasoning24.0% ±4.355 / 90
FrontierMath-Tiers-1-3-v2-PrivateMath46.7% ±3.056 / 90
IOI v1Coding6.8% ±3.339 / 61
ARC-AGI-2 (semi-private)Reasoning4.4%85 / 131
Terminal-Bench 1.0Agentic30.0% ±5.132 / 47
Terminal-Bench 2.0 · Vals AIAgentic27.0% ±4.750 / 66
FrontierMath-Tier-4-v2-PrivateMath12.2% ±5.258 / 72
SWE-Bench verified · Epoch AICoding64.7% ±2.226 / 32
Vibe Code Bench v1.1Coding14.2% ±3.5100 / 123
SimpleQA VerifiedKnowledge21.6% ±1.376 / 91
SWE-benchCoding60.8% ±2.276 / 87
Mystery Game PuzzlesReasoning5.0% ±2.275 / 80
MedQAHealthcare96.1% ±0.25 / 95
CaseLaw v2Legal68.5% ±1.14 / 53
FrontierMath-Tier-4-2025-07-01-PublicMath0.0%3 / 36
TaxEval v2Finance75.2% ±0.822 / 143
MGSMMultilingual92.6% ±1.615 / 74
MortgageTaxFinance66.9% ±0.931 / 97
FrontierMath-2025-02-28-PublicMath30.0% ±15.322 / 64
MedCodeHealthcare43.0% ±2.048 / 106
Finance Agent v1.1Finance51.9% ±2.926 / 51
MedScribeHealthcare80.6% ±1.956 / 109
ARC-AGI-2 (public eval)Reasoning5.8%64 / 122
CorpFin v2Finance60.2% ±1.070 / 131
LegalBenchLegal81.8% ±0.580 / 149
ARC-AGI-1 (public eval)Reasoning61.5%68 / 116
SWE-bench Multilingual · swebench.comMultilingual39.7%12 / 12
SWE-bench Multilingual · SWE-bench/experimentsCoding39.7%14 / 14

Sources

BenchLMCC BY-NC 4.0OpenRouterTerms restrict reuseEpoch AICC BYVals AINo licence statedEuroEvalMITSWE-bench teamNo licence statedARC Prize FoundationTerms restrict reusemodels.devMIT

History

5 releases
Index49.5 → 48.4
23 Oct 202612 Nov 2026
Output / 1M$2.00 → $2.00
23 Oct 202612 Nov 2026

Questions

How much does GPT-5 mini cost?
$0.25 per 1M input tokens, $2.00 per 1M output tokens.
How good is GPT-5 mini?
Its index is 48.4 out of 100. That is place 158 of 283 ranked models.
What is GPT-5 mini best at?
Math: 52.2 out of 100, place 77 of 168.
How fast is GPT-5 mini?
78 output tokens per second, 2.6 s to the first token.
What is the context window of GPT-5 mini?
400K tokens.
Is GPT-5 mini open source?
No. Its weights are not published.
When was GPT-5 mini released?
August 2025.

Badge

GPT-5 mini on Inference360: place 158 of 283, index 48.4
HTML
<a href="https://inference360.ai/models/gpt-5-mini"><img src="https://inference360.ai/badges/gpt-5-mini.svg" alt="GPT-5 mini on Inference360: place 158 of 283, index 48.4" height="20"></a>
Markdown
[![GPT-5 mini on Inference360: place 158 of 283, index 48.4](https://inference360.ai/badges/gpt-5-mini.svg)](https://inference360.ai/models/gpt-5-mini)

Compare with

Kimi K2 Thinking48.2Qwen3.5 Plus Thinking48.1GPT-5 nano41.9

Lower price

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

More from OpenAI

All models
GPT-5 (high)53.6GPT-5 (medium)53.0GPT-5 nano41.9GPT-555.0GPT-OSS 120B43.4