OpenAI · Released Mar 17, 2026GPT-5.4 mini
Strong small GPT for coding subagents, quick tool use, and high-volume work.
- Capability
- 148.8ECI · #56 of 148
- Input
- $0.75per 1M tokens
- Output
- $4.50per 1M tokens
- Context
- 400K128K max output
What it costs
- Input
- $0.75per million tokens
- Output
- $4.50per million tokens
- Cached input
- $0.075per million tokens
- Blended (3:1)
- $1.69pricier than 66% of priced models
Typical monthly bills
| Monthly usage | Estimated cost |
|---|---|
| Side project2M input + 0.5M output tokens | $3.75 |
| Team assistant25M input + 5M output tokens | $41.25 |
| Production app250M input + 50M output tokens | $412.50 |
Official OpenAI API price, as listed on models.dev.
Independent benchmarks
#56 of 148 scored models
The shaded band is Epoch AI’s confidence range (147.1–150.5); the tick marks the median scored model.
Scores from Epoch AI, run independently of OpenAI.
The details
- Lab
- OpenAI
- Released
- Mar 17, 2026
- Knowledge cutoff
- Aug 31, 2025
- Context window
- 400,000 tokens
- Max output
- 128,000 tokens
- Inputs
- Text, Images
- Output
- Text
- Reasoning
- Adjustable effort low · medium · high · xhigh
- Tool calling
- Yes
- Structured output
- Yes
- Weights
- Proprietary
- API model ID
gpt-5.4-minion OpenAI- Availability
- 29 API providerslisted on models.dev
What OpenAI claims
Published by the lab at launch. Settings vary, so compare these only with care.
| Benchmark | Score | Setting | Source |
|---|---|---|---|
| SWE-Bench Pro | 54.4 resolve rate | reasoning effort xhigh | Source |
| Terminal-Bench v2.0 | 60 | reasoning effort xhigh | Source |
| MCP Atlas | 57.7 | reasoning effort xhigh | Source |
| Toolathlon | 42.9 | reasoning effort xhigh | Source |
| τ²-Bench Telecom | 93.4 | reasoning effort xhigh | Source |
| GPQA Diamond | 88 | reasoning effort xhigh | Source |
| Humanity's Last Exam | 41.5 | with tools | Source |
| Humanity's Last Exam | 28.2 | without tools | Source |
| OSWorld-Verified | 72.1 success rate | reasoning effort xhigh | Source |
| MMMU Pro | 78 | with Python | Source |
| MMMU Pro | 76.6 | without tools | Source |
| OmniDocBench v1.5 | 0.1263 overall edit distance | reasoning effort none | Source |
More from OpenAI
About GPT-5.4 mini
How much does GPT-5.4 mini cost?
GPT-5.4 mini costs $0.75 input / $4.50 output per million tokens (official OpenAI API price). Cached input is $0.075 per million tokens. At a 3:1 input-to-output mix that is $1.69 per million tokens, more expensive than 66% of the 360 priced models we track.
What is the context window of GPT-5.4 mini?
GPT-5.4 mini accepts up to 400,000 tokens per request and can write up to 128,000 tokens in one response.
How good is GPT-5.4 mini?
Epoch AI gives GPT-5.4 mini a Capabilities Index score of 148.8 (likely range 147.1–150.5), ranking it #56 of 148 models Epoch has scored. Epoch AI benchmark results: GPQA Diamond 86.9%, FrontierMath Tiers 1–3 51.2%, OTIS Mock AIME 2024–2025 88.9%, SimpleQA Verified 29.4%.
Is GPT-5.4 mini open source?
No. GPT-5.4 mini is proprietary; you use it through OpenAI’s API or partner platforms.
What inputs does GPT-5.4 mini support?
GPT-5.4 mini accepts text and images and replies in text. It is a reasoning model with low, medium, high and xhigh effort settings, supports tool calling and can return structured JSON output.
When was GPT-5.4 mini released?
OpenAI released GPT-5.4 mini on Mar 17, 2026. Its training data runs to Aug 31, 2025.
What are the best alternatives to GPT-5.4 mini?
The closest current models from other labs on capability, price and release date are Grok 4.3 (xAI, ECI 149.2, $1.25 / $2.50), GLM-5.1 (Z.ai (Zhipu), ECI 149.9, $1.40 / $4.40), Kimi K2.7 Code (Moonshot AI, ECI 150.0, $0.95 / $4.00) and Qwen3.6 Plus (Alibaba (Qwen), ECI 147.6, $0.50 / $3.00).