Skip to content
Longterm Wiki

GPT-4o mini

OpenAI

GPT-4o mini was released July 18, 2024 as the most cost-efficient small model from OpenAI. Priced at \$0.15/\$0.60 per million tokens — over 60% cheaper than GPT-3.5 Turbo. Scored 82% on MMLU and 87.2% on HumanEval. Replaced GPT-3.5 Turbo as the default free-tier model in ChatGPT.

Source →
Developer
OpenAI
Released
2024-07-18
Context Window
128K tokens

Pricing

TypePrice per MTok
Input$0.15
Output$0.60

Benchmarks6

General

1 benchmark
IFEval
80.6%#9/10
15th percentile

Knowledge

2 benchmarks
MMLU
82%#30/37
19th percentile
MMLU-Pro
63.2%#15/15
3th percentile

Reasoning

1 benchmark
GPQA Diamond
39.8%#31/34
10th percentile

Math

1 benchmark
MATH
70.2%#24/31
24th percentile

Coding

1 benchmark
HumanEval
87.2%#16/25
38th percentile
Percentile among tested models:Top 25%50-75%25-50%Bottom 25%

GPT Family7

ModelTierReleasedInput $/MTok
GPT-4.1flagship2025-04-14$2.00
GPT-4.1 minimini2025-04-14$0.40
GPT-4.1 nanonano2025-04-14$0.10
GPT-4oflagship2024-05-13$2.50
GPT-4 Turboflagship2023-11-06$10.00
GPT-4flagship2023-03-14$30.00
GPT-3.5 Turbomid2023-03-01$0.50

Other OpenAI Models6

o32025-04-16
o-series (flagship)
o4-mini2025-04-16
o-series (mini)
o3-mini2025-01-31
o-series (mini)
o12024-12-05
o-series (flagship)
o1-preview2024-09-12
o-series (preview)
o1-mini2024-09-12
o-series (mini)

Details

Model FamilyGPT
Tiermini
Generation4o-mini
Release Date2024-07-18
Context Window128K tokens
Training Cutoff2023-10
Open WeightNo
Modalitytext, image

Capabilities2

tool-usevision

Sources1

GPT-4o mini

Tags

gptgpt-4ocost-efficient