Changelog
New models, pricing and billing rules, API behavior and docs updates, listed by date.
- The website is now available in Japanese, Korean, French, German and Spanish. Docs remain in English and Simplified Chinese.
- Billing fix: OpenAI streaming responses that end without any output are no longer billed.
- Listed Gemini 4 Argon, Gemini 4 Carbon, Kimi K3, Kimi K3.1 Pro, Hy4, MiniMax M3 and MiniMax M3.1 Flash.
- Every model now has its own page with prices, 24-hour status and setup details; the pricing page groups models by vendor.
- Docs moved to docs.flyrouter.ai. Old links redirect automatically.
- GET /v1/models lists models without an API key.
- Added notification emails (low balance, top-up received) and automatic top-up with a saved card.
- Availability on the pricing page, playground and status page now shows real success rates; the status page shows availability by hour.
- Requests for a model that does not exist now return 404.
- When a client aborts a streaming request, only the output already delivered is billed.
- Listed GPT Image 2.6 Flare and GPT Image 2.6 Sunburst.
- The pricing page shows 1-hour cache write prices and tool surcharges.
- Fixed Codex namespace and freeform tools being dropped when requests were converted.
- The API reference is now interactive, so you can send requests from the page.
- Added account management APIs: check balance, manage API keys and read usage with a management token.
- The pricing page shows cache write prices.
- Docs are now split into separate pages, each with its own URL.
- The website launched in English and Traditional Chinese.
- The playground includes free image generations every day.
- Added the playground: sign in to try chat and image models directly.