What is o4‑mini?
o4‑mini is OpenAI’s newest small‑footprint reasoning model. It builds on o3‑mini with markedly better performance in coding, math, science, and everyday problem‑solving, while adding multimodal and tool capabilities previously limited to larger models.
Key improvements over o3‑mini include:
Access to advanced ChatGPT tools, including browsing, Python (code interpreter), file uploads, vision inputs, image generation, and memory
Faster average latency and higher throughput ceilings
128k‑token context window (same as o3) for longer conversations and documents
How is o4‑mini different from o3 and o3‑mini?
o4‑mini delivers faster responses than o3‑mini with a larger context window and is comparable to o3’s accuracy with higher reasoning settings. In addition, o4‑mini inherits algorithmic improvements from o3, giving noticeably better results on programming and quantitative tasks compared with o3‑mini. Finally, o4-mini has access to an expanded set of tools and tool use compared to its predecessor.
Which ChatGPT plans have access to o4‑mini?
Enterprise & Edu: Available starting April 24, 2024.
Plus, Team & Pro: All these plans can select o4‑mini directly in the model picker (Pro includes expanded, near‑unlimited usage limits).
Free: o4‑mini will replace o3‑mini gradually over the coming weeks; availability may fluctuate as we scale capacity.
Does o4‑mini replace o3‑mini?
Yes. Going forward, o4-mini is the recommended small reasoning model in the o-series, offering improved performance, faster responses, expanded agentic tool use, and a range of reasoning modes.
Tools available with o4‑mini
Browsing – search the live web for up‑to‑date information and cite sources.
Python (Code Interpreter) – run Python code to analyze data, create charts, and transform files.
File uploads & data analysis – upload documents, spreadsheets, and images for deep analysis directly in ChatGPT.
Vision inputs – understand and reason about images you provide.
Image generation (DALL·E) – create original images from text prompts inside ChatGPT.
Memory & custom instructions – let ChatGPT remember preferences across sessions when enabled.
Voice (mobile) – converse hands‑free using voice on the ChatGPT mobile app.
Can I use o4‑mini through the API?
Yes. o4‑mini is available in the Chat Completions and Responses API with streaming, JSON mode, and function calling. Developers in usage tiers 1 to 5 can access the model immediately. See usage tiers for token‑per‑minute limits.
Does o4‑mini support vision or image inputs?
Yes. Unlike o3‑mini, o4‑mini can analyze images you upload and accept screenshots or photos in the prompt. It also supports generating images with DALL·E.
How does o4‑mini ensure safe and reliable outputs?
All outputs are governed by the same multilayer safety system used for GPT‑4o and o3. This includes real‑time policy filters, extensive red‑teaming, and continuous post‑deployment monitoring.
What usage limits will be enforced on o4‑mini?
Plan | Daily messages (o4‑mini) |
Plus & Team & Enterprise & Edu | 300 (100 for o4-mini-high) |
Pro | Near‑unlimited (subject to automated safeguards) |
Free | Limited pool replenishes every 5h |
Is fine‑tuning available for o4‑mini?
Yes, fine‑tuning capabilities are available for o4‑mini.
Where can I read the latest release notes?
See the Model Release Notes for details on each update.