Skip to main content
​
​

Product

  • Model pages, provider pages and the models list label audio models as Speech, Transcription and Music instead of TTS and STT (#417).
  • Model cards show prices for image models such as Grok Imagine Image; before, some of them appeared as free (#418).
​

API & SDKs

  • The TypeScript, Python, Go, C#, Java, PHP and Ruby SDKs 2.0.1 accept any model ID string, so new models work without waiting for an SDK release (#367, #393).
  • Generated SDK model constants now list only models the gateway can call, not every catalogue entry (#367).
​

Product

  • Low-balance emails are sent when a workspace’s balance drops below the threshold set in Settings -> Credits -> Low balance email alert (#370, #371).
  • Model pages list six providers before collapsing the list, and say when a catalogued model has no API provider yet (#372).

API & SDKs

  • Requests on a non-standard pricing plan such as priority are billed at standard rates for usage that plan doesn’t price; before, that usage went unbilled (#374).
​

API & SDKs

  • Gateway responses report latency_ms as the time to the first frame, separate from generation_ms, including for non-streamed requests (#360).
​

Product

  • Credits use one Standard plan with a flat 5% top-up fee, replacing the Basic and Enterprise tiers; see Pricing Tiers (#351).
  • Every account gets a personal workspace on first sign-in, paid accounts can create more, and new API keys go to the workspace selected in the header; see Workspaces (#352, #351).
  • The model page playground no longer fails with invalid_secret after an API key rotation (#355).

API & SDKs

  • Image requests to openai/gpt-image-2 no longer fail for missing pricing when a price exists (#355).
​

Product

  • The model page playground adds TTS and Music modes for models that generate speech or music (#348).

API & SDKs

  • Breaking: The SDKs 2.0.0 remove the qwen/qwen3-tts-tokenizer-12hz model ID; update any code that references it (#337).
​

Product

  • Teams are now workspaces, and members, invites, join requests and roles are managed in Settings -> Workspaces; see Workspaces (#336, #345).
  • The new Announcements page lists Phaseo announcements, and the homepage shows the latest three (#335, #339).
​

API & SDKs

  • The SDKs 1.2.0 add the x-ai/grok-4.3 model ID (#314).
​

Product

  • The minimum manual credit top-up is now $5, and auto top-up still starts at $1 (#315).
​

Product

  • Model pages add a Pricing tab with the past hour’s traffic-weighted effective price and each provider’s pricing history (#305, #307).
  • The model page playground adds image, video, audio, embeddings and moderation modes (#307).
  • The Apps section on model pages lists public apps seen in gateway traffic, with their request and token counts (#302).

API & SDKs

  • Claude Opus 4.7 requests on Anthropic, Google Vertex and Amazon Bedrock accept xhigh reasoning effort, and the gateway drops temperature, top_p and top_k; see Anthropic Claude Opus 4.7 (#308).
​

Product

  • The release calendar has a page per organisation showing its releases on today’s weekday or all its releases, and now covers every past release (#300).

API & SDKs

  • Requests to free models, such as :free IDs, work with no credits; paid requests still need at least $1 in credits (#293, #294).
  • Providers without configured pricing are skipped, and a request with no priced provider returns unsupported_model_or_endpoint with reason pricing_not_configured (#293).
  • Breaking: Requests to announced, withheld or retired models, or models past their deprecation date, are now rejected, and SDK validateModel returns ok: false; switch to an active model (#298).
​

Product

  • The Chat audio room generates music from a prompt and lyrics, and keeps generated tracks in your browser (#277).
  • Usage and Logs show each model’s organisation-prefixed Phaseo name instead of the provider’s name (#277).
  • Model cards and the models table show names as Organisation: Model, for example Moonshot: Kimi K2.6 (#278).
​
​

Product

  • Compare pages show each model’s gateway usage and display all benchmark scores in the same format (#244).
  • Provider pages show usage trends for each model they serve and app usage over time (#247).
  • The release calendar at /updates/calendar shows which weekdays each organisation tends to release on and the chance of a release today (#264).
​

Product

  • Model pages mark models as announced or withheld and hide the provider, app and performance sections that don’t apply to them (#245, #259).
  • The website adds step-by-step guides for moving from other AI gateways at /migrate (#248).

API & SDKs

  • POST /v1/rerank ranks documents by relevance to a query, with support in the TypeScript and Python SDKs; see Rerank documents (#242).
​
​

Product

  • The Pay As You Go top-up fee drops from 7% to 5%, and the $1 minimum fee is unchanged (#225).
  • Cards for image, video and audio models show a From: price in each model’s own unit instead of a token price (#225).
  • Analytics and session recording on phaseo.app run only after you accept analytics cookies (#226).

API & SDKs

  • Chat Completions responses include token usage from providers that report input_tokens and output_tokens instead of prompt_tokens and completion_tokens (#227).
​

Product

  • Model pages on phones stack pricing in two columns and keep the Other models row within the screen (#223).
  • Providers that serve a model for free appear under the Free pricing tier instead of Standard (#223).
​
Last modified on October 28, 2026