What it is
This page discusses OpenAI’s new o1 chain-of-thought models, highlighting the release of two models, o1-preview and o1-mini, and their implications in terms of cost, performance, and reasoning capabilities.
Gabriel’s notes
OpenAI released two major new preview models today: o1-preview and o1-mini (that mini one is not a preview)previously rumored as having the codename strawberry. There’s a lot to understand about these modelsthey’re not as simple as the next step up from GPT-4o, instead introducing some major trade-offs in terms of cost and performance in exchange for improved reasoning capabilities.
Good fit if you want to:
- go deeper on technical details, benchmarks, or model/system behavior.
Pricing snapshot (auto-enriched): No free tier; usage-based pricing requiring at least $1,000 spent on API credits; hidden reasoning tokens are billed but not visible in API responses.
Work-use / compliance snapshot (auto-enriched): OpenAI’s o1 models, accessible via the API platform, are suitable for workplace use with enterprise-grade data ownership and control, no default training on customer data, configurable data retention, SAML SSO support, and compliance with SOC 2 Type 2, HIPAA (via BAA), GDPR, and other industry standards.
Alternatives (auto-enriched): Alternative: DeepSeek R1 | Comparison: DeepSeek R1 delivers comparable reasoning performance to OpenAI’s o1 models but is significantly more cost-efficient, using far fewer GPUs and offering much lower pricing. Alternative: OpenAI GPT-4o | Comparison: GPT-4o supports image inputs and faster response times, making it better suited for applications needing these features,…
Reading tip: skim headings first, then focus on the sections that match your current project or question.
Author: Simon Willison
Note: pricing and policy details can change—verify on the official site before making decisions.