Today we’re releasing Orchid 2, the most careful model we have ever trained. It reads up to a million tokens in a single prompt — roughly 2,500 pages — and keeps every one of them in view while it reasons.
What’s new
Deep thinking is now 40% faster to its first token, citations are available on every plan, and Studio can compare up to four prompts side by side. Existing API keys work without any changes.
Built on what you told us
The most common request over the last year was simple: let me give Orchid everything and trust it to find what matters. Orchid 2 is our answer to that request.
