Model

GPT-4o mini

CLOSED

OpenAI · GPT-4o family · released Jul 18, 2024

Small, cheap GPT-4o sibling that replaced GPT-3.5 Turbo as OpenAI's default budget model in ChatGPT and the API.

ReasoningCodingVisionFunction callingTool useAgentic

973.1

Elo · rank #220

Parameters

Undisclosed

Active params

—

Context

128K tokens

Architecture

Dense, multimodal Transformer, small/low-cost tier (architecture/size undisclosed)

License

Proprietary

Languages

—

API price (in/out)

$0.15 / $0.6

Modalities

text · vision

Benchmark results

Bar shows position within the tracked field; marker = field best

ARC-AGI-1Reasoning3.6%#18

best: GPT-5.6 · 96.5%

ARC-ChallengeReasoning93.3%#14

best: Llama 3.1 405B · 96.9%

ARC-EasyReasoning96.8%#3

best: Phi-3-medium (14B) · 97.7%

BIG-Bench HardReasoning77.7%#39

best: ERNIE 4.5 300B-A47B · 94.3%

BigCodeBenchCoding57.4%#1

best: this model · 57.4%

DROPReasoning79.7#26

best: Hunyuan-T1 · 93.1

GPQA DiamondReasoning40.2%#217

best: GPT-5.6 · 94.6%

GSM8KMath94.1%#18

best: Llama 3.1 405B · 96.8%

HumanEvalCoding87.2%#35

best: Claude Opus 4.5 · 99.4%

MATH-500Math70.2%#90

best: GPT-5 · 99.4%

MGSMMath87.0%#19

best: OpenAI o4-mini · 93.7%

MMLU-ProKnowledge63.1%#111

best: Claude Fable 5 · 91.5%

MMLUKnowledge82.0%#76

best: OpenAI o3 · 92.9%

MMMUVision59.4%#80

best: Claude Fable 5 · 89.3%

PIQAReasoning93.1%#1

best: this model · 93.1%

PutnamBenchMath0.0%#9

best: DeepSeek-Prover-V2 671B · 7.1%

SimpleQAKnowledge9.5%#39

best: GPT-4.5 · 62.5%

Social IQaReasoning79.0%#4

best: Apple DCLM-Baseline 7B · 82.9%

WinoGrandeReasoning79.2%#41

best: PaLM 2 · 90.9%

Run it locally

Closed weights — available via API only. No local deployment.

Input / M tok

$0.15

Output / M tok

$0.6

GPT-4o family

Elo progression across releases

GPT-4oMay 20241249.9 GPT-4o miniJul 2024973.1

Resources

DOCSdevelopers.openai.com/api/docs/models/gpt-4o-mini APIAvailable via OpenAI API

API price $0.15/$0.6 · each benchmark row carries its own source badge (see methodology)