Skip to main content
​

API & SDKs

  • Breaking: Text requests set the tier only with service_tier; replace speed: "fast" with service_tier: "priority", and synchronous service_tier: "batch" requests now return a validation error; see Service Tiers (#565).
​

Product

  • Model page quickstarts list the endpoints a model supports with their HTTP methods, and choosing one updates the example code (#521).

API & SDKs

  • Requests with service_tier: "priority" now route only to providers that support that tier, and Venice serves Claude Opus 4.8 priority requests under the same model ID; see Service Tiers (#524, #530).
​

Product

  • Model pricing panels keep the plan selector visible when a provider has only one active plan, such as batch-only pricing (#515).
  • Anthropic on AWS routes show the AWS logo and clearer provider names in model pricing and provider lists (#519, #525, #526).
​
​
​

Product

  • Models lists the free router, and its page shows which free models it can route to and recent routed usage; see Launch on the free router (#495).
​
​
​
​

Product

  • Chat can retry a failed image or video generation with the same prompt, model and temporary chat setting (#479).
​

API & SDKs

  • The TypeScript, Python, Go, C#, Java, PHP and Ruby SDKs 2.0.4 add session_id, structured error diagnostics and model availability fields to their types; see release notes (#476).
​

Product

  • Pricing hover cards on Models show image, video and tiered text prices, grouped by quality, resolution, duration, audio and context size (#474).
  • Models shows minute-billed audio models such as whisper-1 at their per-minute price, for example $0.006 per minute, instead of a rounded token price (#477).
  • Model page headers add a model ID control that switches between the model’s aliases and copies the selected ID (#473).
  • Filter counts in the Models sidebar now update to match the other filters you select (#473).

API & SDKs

  • Breaking: /v1/videos, /v1/batches, /v1/files and /v1/music/generations are temporarily disabled and return 501 not_implemented; pause calls to them until they are re-enabled (#473).
​

API & SDKs

  • The List models endpoint accepts availability=all to include coming-soon and inactive models, and each provider entry returns availability_status and availability_reason; see List models (#466).
  • Error responses include failed_providers, routing_diagnostics and provider_failure_diagnostics, which point to causes such as missing credentials or region limits; see Errors and Debugging (#467).
  • The Get a generation endpoint returns replay_request and replay_supported, so a logged request can be sent again to its original endpoint; see Get a generation (#467).
​
​

Product

  • Settings -> Usage splits activity into Logs, Jobs and Sessions views, with live refresh and filters and time ranges kept in the URL (#463, #467).
  • The model playground shows Retry and Delete on failed image and video generations, and Retry reuses the same prompt and model (#465).

API & SDKs

  • Chat Completions and Responses requests accept session_id, which groups related requests into one session in usage logs (#463).
​

API & SDKs

  • Workspace policies that block models or providers are enforced again, so blocked requests are rejected before routing (#444).
​

Product

  • The model playground adds TTS and Music modes for text-to-speech and music generation models (#441).
  • Model cards on Models no longer show Free for paid models, and the coming-soon filter count matches the models listed (#441).

API & SDKs

  • The TypeScript, Python, Go, C#, Java, PHP and Ruby SDKs 2.0.3 add the latest model and benchmark IDs to their typed lists; see release notes (#439).
Last modified on October 28, 2026