OpenAI o1 reasoning-excelling model is now available as an API for developers

Good news for developers. OpenAI’s o1, which excels in reasoning, is now available in API.

The Microsoft-backed AI company announces that OpenAI o1’s API features function calling, structured outputs, developer messages, and vision features. And, we’re also getting real-time updates like WebRTC support to integrate it across platforms.

The o1 series of models, which include the full o1 and the more affordable o1-mini, are designed for reasoning tasks. It’s currently available to Tier 5 in the API, the most expensive tier of them all, which is priced at $1,000 with a usage limit of $200,000 per month.

The o1 model handles both text and images and produces structured text outputs, while the o1-mini is more geared towards specialized text tasks. Both have a knowledge cutoff of October 2023, with the latest versions offering extended context windows and varying token limits.

The o1-2024-12-17 supports up to 200,000 tokens, while the o1-mini-2024-09-12 supports up to 128,000 tokens.

OpenAI also mentions that it’s launching a “preference fine-tuning method”, which lets you customize model behavior based on user and developer preferences. There are also new SDKs for Go and Java in beta, and it’s also cutting prices across the Realtime API, including for GPT-4o and GPT-4o mini.

In another API-related news, OpenAI also said that it has no plans to release Sora’s API any time soon. The video-generating model arrived for all users not too long ago after months of delay and a beta-testing period. Though, there may change due to popular demands.