I am excited to be introducing GPT-4.1 in the OpenAI API — a major upgrade that redefines what's possible for developers building with AI.
Here’s what’s new and noteworthy:
🧩 3 powerful models:
GPT-4.1: State-of-the-art performance across coding, long-context, and instruction following
GPT-4.1 mini: Matches GPT-4o in intelligence with half the latency and 83% lower cost
GPT-4.1 nano: Fastest, most efficient model for tasks like classification and autocompletion
🧠 Performance highlights:
54.6% on SWE-bench Verified (21.4% better than GPT-4o)
38.3% on MultiChallenge for complex instruction following
1M token context — ideal for massive codebases or document processing
Better vision & image understanding
Enhanced tool use & code diffs for real-world dev workflows
💡 Use cases exploding: From intelligent agents and legal document review to complex frontend builds, this release is built for real-world needs — all with lower cost, reduced latency, and smarter task handling.
🎯 Only available via the API, not ChatGPT — perfect for devs, startups, and enterprises ready to scale serious applications.
Let us know what you’re building with it — the possibilities just hit a whole new level!
Fantastic update! I really appreciate how much more cost-efficient it’s become — that’s a huge win for indie makers and small teams like mine.
I’ll be using this in my next project. Keep up the awesome work!
🎉 coding accuracy improvements is very useful!
Curious about what does the Better vision & image understanding mean? When will the 4o model API public? I mean the text-to-image feature.
Great update and I love how cost efficient it became! will be using this for my next project 100%
Instruction following was a huge improvement for us, it can now reliably output html, json better when instructed instead of defaulting to Markdown!
A fresh lineup of GPT models brings major upgrades in coding, instruction handling, and long-form content understanding. Now built to follow directions better, write cleaner code, and keep track of much longer conversations or documents. Also introducing the first-ever nano model—compact, fast, and built to run smoothly on your own device.
@sama Massive upgrade! The latency and cost reductions alone make GPT-4.1 Mini a game-changer — and 1M token context? Perfect for handling real codebases. Can’t wait to build with this! 👏🔥
