OpenAI has launched GPT-4.1, a successor to the GPT-4o multimodal AI mannequin launched by the corporate final yr. Throughout a livestream on Monday, OpenAI stated GPT-4.1 has an excellent bigger context window and is best than GPT-4o in “nearly each dimension,” with massive enhancements to coding and instruction following.
GPT-4.1 is now obtainable to builders, together with two smaller mannequin variations. That features GPT-4.1 Mini, which, like its predecessor, is extra reasonably priced for builders to tinker with, and GPT-4.1 Nano, an much more light-weight mannequin that OpenAI says is its “smallest, quickest, and most cost-effective” one but.
All three fashions can course of as much as a million tokens of context — the textual content, photographs, or movies included in a immediate. That’s excess of GPT-4o’s 128,000-token restrict. “We educated GPT‑4.1 to reliably attend to data throughout the total 1 million context size,” OpenAI says in a put up asserting the fashions. “We’ve additionally educated it to be way more dependable than GPT‑4o at noticing related textual content, and ignoring distractors throughout lengthy and brief context lengths.”
The launch comes as OpenAI plans to part out its two-year-old GPT-4 mannequin from ChatGPT on April thirtieth, asserting in a changelog that current upgrades to GPT‑4o make it a “pure successor” to interchange it. OpenAI additionally plans to deprectate the GPT-4.5 preview within the API on July 14th, as “GPT‑4.1 provides improved or comparable efficiency on many key capabilities at a lot decrease value and latency. ”
OpenAI can be set to debut the total model of its o3 reasoning mannequin and an o4 mini reasoning mannequin any day now, with references having already been noticed within the newest ChatGPT net launch by AI engineer Tibor Blaho.