MODELS
GPT-5 mini
A cheaper GPT-5 tuned for high-volume tool use and everyday agent work.
Specs
- Context window
- 200,000
- Max output
- 16,384
- Modalities
- text, image
- Tool use
- ✓
- Vision
- ✓
- Streaming
- ✓
- License
- proprietary
- Released
- 2025-08-01
Pricing
- Input / 1M
- $0.50
- Output / 1M
- $2.00
- Cached input / 1M
- $0.05
Cost estimate
GPT-5 mini is OpenAI's smaller, cheaper sibling to GPT-5, sharing the same 200K context, vision input, and tool-calling behavior at $0.5/$2 per million tokens. It's aimed at builders who need GPT-5-style reasoning and function calling at roughly a quarter of the cost — chatbots, classification, RAG pipelines, and agent loops where you'll burn through a lot of tokens. You give up some depth on hard reasoning and long-horizon coding compared to full GPT-5.
Editor's verdict
Pick GPT-5 mini when you've already prototyped on GPT-5 and the bill is the problem — it keeps the tool-calling reliability and vision support that make GPT-5 nice to build against, at a price closer to Gemini 2.5 Flash and Claude Haiku. For genuinely hard reasoning, multi-step coding, or anything where one wrong call is expensive, stay on full GPT-5. It's a default workhorse, not a frontier model.
Reviews
No reviews yet. Be the first.
Last updated: 2026-04-29