Changelog

New models, features and fixes — updated as we ship.

September 2026

  • New ModelGPT-6 Astra and GPT-6 Astra Pro are live — OpenAI's new flagship line at 1.05M context.
  • New ModelClaude Fable 5.1 added — Anthropic's frontier model with 1M token context.
  • New ModelGemini 3.8 Flash, Qwen3.8 Max, Muse Spark 1.3, DeepSeek V4.1 Flash and Fugu Ultra v2 now available.
  • ImprovementFailover switching time reduced from ~800ms to under 300ms across all regions.

August 2026

  • New ModelGrok 4.6, GLM 5.3, DeepSeek V4 Pro (0813), Hunyuan 4 Preview and Seed 2.1 Turbo added.
  • FeaturePer-key usage limits and per-key cost breakdown in the console.
  • FeatureModel routing hints (:nitro / :floor) are now generally available.
  • FixResolved an issue where streaming responses could double-count the final chunk's tokens.

July 2026

  • New ModelClaude Opus 5, Claude Sonnet 5, Kimi K3 and the GPT-5.6 family (Sol / Terra / Luna) are live.
  • FeatureCSV export for monthly statements in Billing.
  • ImprovementP50 first-token latency improved 18% on Anthropic routes via a new UAE South region.

June 2026

  • New ModelMistral Medium 3.5, GLM 5.2, MiniMax M3 and Kimi K2.7 Code added to the catalog.
  • FeaturePlayground now supports per-model parameter presets and streaming output.
  • ImprovementCatalog search now matches provider aliases and capability tags.