GPT-4o mini

GPT-4o mini

Quick answer: GPT-4o mini is OpenAI's cost-efficient multimodal model, released July 18, 2024 (API: gpt-4o-mini-2024-07-18). It scores 27.5% on LiveCodeBench and supports text and vision input with a 128K-token context window. At $0.15/$0.60 per 1M tokens, it was the most affordable OpenAI model with vision capability at launch.

At a Glance

Where GPT-4o mini leads

• $0.15/$0.60 per 1M tokens — highly cost-efficient for high-volume pipelines
• Vision support at the mini tier — processes images without the cost of GPT-4o
• 128K-token context window
• Fast latency optimised for real-time applications
• Strong instruction-following from the GPT-4o training lineage

Where it lags

• 27.5% LiveCodeBench — well below GPT-4.1 (not shown) and GPT-4o
• Superseded by newer cost-tier models (GPT-5 mini, o4-mini) for most tasks
• Knowledge cutoff October 2023 — outdated for time-sensitive queries
• No reasoning/extended-thinking mode

Best for: High-volume classification, content moderation, basic code completion, image captioning, and legacy integrations built on the GPT-4o mini API.

What GPT-4o mini Is

GPT-4o mini is the small-model tier of OpenAI's GPT-4o family. Released in July 2024 to replace GPT-3.5 Turbo as the default cost-efficient option, it introduced multimodal vision capability at a price point ($0.15/$0.60 per 1M) that made image-understanding accessible for high-volume production workloads.

The model's architecture shares the same training approach as GPT-4o — instruction fine-tuning with RLHF — at reduced scale. This gives it GPT-4o-class instruction following and safety at lower capability ceiling and significantly lower cost and latency.

While GPT-4o mini has largely been superseded by GPT-5 mini and o4-mini for new deployments (both offering better performance at similar or lower price), it remains widely used in existing integrations and is a stable, well-tested choice for teams that don't need updated knowledge or advanced reasoning.

Specifications

Pricing

Prompt caching provides a 50% discount on cached input. Pricing per OpenAI pricing page.

Context Window

GPT-4o mini has a 128,000-token context window — roughly 90 pages of text, sufficient for most document-analysis and multi-turn conversation tasks.

Public Benchmark Scores

Additional well-known scores at launch (OpenAI technical report, July 2024): MMLU 82.0%, MATH 70.2%, HumanEval 87.2%.

GPT-4o mini vs Alternatives

At identical pricing to Gemini 2.5 Flash ($0.15/$0.60), GPT-4o mini offers a much smaller context window (128K vs 1M) and lower coding benchmark scores. For new cost-tier deployments, Gemini 2.5 Flash or GPT-5 mini are the recommended alternatives.

Use GPT-4o mini via API

from openai import OpenAI

client = OpenAI(api_key="YOUR_API_KEY")

response = client.chat.completions.create(
    model="gpt-4o-mini",
    messages=[{"role": "user", "content": "Classify this text as positive, negative, or neutral..."}],
)
print(response.choices[0].message.content)

Related Models & Benchmarks

• Compare: GPT-4o · GPT-5 mini · Gemini 2.5 Flash · Claude Haiku 3.5
• Benchmarks: LiveCodeBench

Frequently Asked Questions

Specs and scores sourced from OpenAI's official GPT-4o mini announcement (July 2024) and Benchgen evaluations. Pricing cited to the OpenAI pricing page. Last updated 2026-07-24.