# GPT-5.4 mini

> OpenAI's fast, low-cost small model that nearly matches GPT-5.4 on coding and computer use.

- Maker: OpenAI
- Line: GPT Mini
- Canonical page: https://ai-tldr.dev/models/gpt-5-4-mini/
- Updated: 2026-06-18

## In short

GPT-5.4 mini is OpenAI's small-tier model in the GPT-5.4 family, released on March 17, 2026 alongside GPT-5.4 nano. It brings much of the capability of the flagship GPT-5.4 to a faster, cheaper model aimed at high-volume work. It has a 400K-token context window and costs $0.75 per million input tokens and $4.50 per million output tokens. You can use it through OpenAI and OpenRouter.

## Facts

| Fact | Value | Source |
| --- | --- | --- |
| Context window | 400K (max output 128K) | [1] |
| Input price | $0.75 / 1M tokens (cached $0.075 / 1M tokens) | [2] |
| Output price | $4.50 / 1M tokens | [2] |
| Licence | Proprietary · API only | [1] |
| Released | 17 March 2026 | [3] |
| Parameters | Undisclosed | [1] |
| Knowledge cutoff | August 31, 2025 | [1] |

## Sources

1. [GPT-5.4 (Wikipedia)](https://en.wikipedia.org/wiki/GPT-5.4)
2. [GPT-5.4 mini model docs (OpenAI API)](https://developers.openai.com/api/docs/models/gpt-5.4-mini)
3. [Introducing GPT-5.4 mini and nano (OpenAI)](https://openai.com/index/introducing-gpt-5-4-mini-and-nano/)

## Overview

GPT-5.4 mini is OpenAI's small-tier model in the GPT-5.4 family, released on March 17, 2026 alongside GPT-5.4 nano. It brings much of the capability of the flagship GPT-5.4 to a faster, cheaper model aimed at high-volume work. OpenAI describes it as its most capable mini model yet, with improvements over the previous GPT-5 mini across coding, reasoning, multimodal understanding, and tool use while running more than 2x faster.

The model has a 400K-token context window and supports up to 128K output tokens, with a knowledge cutoff of August 31, 2025. In the API it accepts text and image input and returns text. It supports function calling, structured outputs, streaming, web search, file search, and computer use, which makes it well suited to agents and subagents that call tools and operate a browser or terminal. Fine-tuning is not offered for this model.

GPT-5.4 mini is available through the OpenAI API (model IDs gpt-5.4-mini and the snapshot gpt-5.4-mini-2026-03-17), inside Codex, and in ChatGPT, where free and Go users can reach it through the Thinking option. It is also generally available in GitHub Copilot. Pricing is $0.75 per 1M input tokens and $4.50 per 1M output tokens, with cached input at $0.075 per 1M, making it one of OpenAI's lowest-cost paths to near-flagship coding quality.

## Benchmarks

- SWE-Bench Pro: 54.4% ([source](https://www.datacamp.com/blog/gpt-5-4-mini-nano))
- OSWorld-Verified: 72.1% ([source](https://www.datacamp.com/blog/gpt-5-4-mini-nano))
- Terminal-Bench 2.0: 60% ([source](https://www.datacamp.com/blog/gpt-5-4-mini-nano))

## Pricing

- Input: $0.75 / 1M tokens per 1M tokens
- Cached input: $0.075 / 1M tokens per 1M tokens
- Output: $4.50 / 1M tokens per 1M tokens

API pricing for gpt-5.4-mini. Text and image input; text output.

Source: https://developers.openai.com/api/docs/models/gpt-5.4-mini

## Strengths

- Near-flagship coding: scores about 54.4% on SWE-Bench Pro, only a few points behind the full GPT-5.4, at a fraction of the cost
- Strong computer use and agentic tool calling (web search, file search, computer use), built for subagent workloads
- Large 400K-token context window with up to 128K output tokens
- More than 2x faster than the previous GPT-5 mini
- Low price: $0.75 input / $4.50 output per 1M tokens, with $0.075 cached input
- Available on the ChatGPT free tier, in the API, in Codex, and in GitHub Copilot

## Best for

- High-volume coding assistants and codebase exploration with grep-style tools
- Agentic and subagent pipelines that call tools, search, and operate a computer
- Real-time image reasoning and multimodal understanding from text plus images
- Cost-sensitive production chat and automation at scale
- Drafting, summarizing, and classifying over long documents using the 400K context window

## How to access

- [OpenAI](https://developers.openai.com/api/docs/models/gpt-5.4-mini) — model id `gpt-5.4-mini`
- [OpenAI](https://developers.openai.com/api/docs/models/gpt-5.4-mini) — model id `gpt-5.4-mini-2026-03-17`
- [OpenRouter](https://openrouter.ai/openai/gpt-5.4-mini) — model id `openai/gpt-5.4-mini`

## GPT Mini versions

- GPT-5.4 mini (17 March 2026) — this page
- [GPT-5 mini](https://ai-tldr.dev/models/gpt-5-mini/) (7 August 2025)
- [GPT-4o mini](https://ai-tldr.dev/models/gpt-4o-mini/) (18 July 2024)

## FAQ

### When was GPT-5.4 mini released and who makes it?

OpenAI released GPT-5.4 mini on March 17, 2026, alongside GPT-5.4 nano. It is the small-tier model in the GPT-5.4 family.

### How much does GPT-5.4 mini cost?

API pricing is $0.75 per 1M input tokens and $4.50 per 1M output tokens, with cached input at $0.075 per 1M tokens.

### What is the context window and what inputs does GPT-5.4 mini accept?

It has a 400K-token context window and supports up to 128K output tokens. In the API it accepts text and image input and returns text; audio and video are not supported.

### How good is GPT-5.4 mini at coding?

It scores about 54.4% on SWE-Bench Pro, only a few points behind the full GPT-5.4, and it also performs strongly on agentic benchmarks such as OSWorld-Verified (72.1%) and Terminal-Bench 2.0 (60.0%).

## Links

- [Introducing GPT-5.4 mini and nano (OpenAI)](https://openai.com/index/introducing-gpt-5-4-mini-and-nano/)
- [GPT-5.4 mini model docs (OpenAI API)](https://developers.openai.com/api/docs/models/gpt-5.4-mini)
- [GPT-5.4 mini on OpenRouter](https://openrouter.ai/openai/gpt-5.4-mini)
- [GPT-5.4 mini and nano: Benchmarks (DataCamp)](https://www.datacamp.com/blog/gpt-5-4-mini-nano)
- [GPT-5.4 mini now GA in GitHub Copilot](https://github.blog/changelog/2026-03-17-gpt-5-4-mini-is-now-generally-available-for-github-copilot/)
- [GPT-5.4 mini hits 54% SWE-Bench Pro on the free tier (Digital Applied)](https://www.digitalapplied.com/blog/gpt-5-4-mini-free-tier-54-swe-bench-pro-performance)
- [GPT-5.4 (Wikipedia)](https://en.wikipedia.org/wiki/GPT-5.4)

Source: AI/TLDR LLM registry, https://ai-tldr.dev/models/gpt-5-4-mini/