What's new:
- Rollout capacity budgets: tg endpoints rollout takes two flags that set a rollout's capacity budgets . --max-surge caps how many replicas each canary step adds to the target. --max-unavailable lets a rollout start from a DEGRADED source, or shut down source replicas to make room for the target.
Also shipped Nov 16, 2026Together AI in November 2026 →
Deferred tier for asynchronous model calls on the Interactions API
Google Cloud announced general availability for the Deferred tier on the Gemini Enterprise Agent Platform. It offers a 50% discount compared to Standard PayGo rates for asynchronous Gemini model calls, providing a 24-hour completion window and 7-day state retention.
DeepSeek-V4.1-Flash-1017 is available in Public Preview
DeepSeek-V4.1-Flash-1017 is available in public preview as a fully managed model in Vertex AI Model Garden. The updated model features improved reasoning, coding, agentic, and visual capabilities, supporting multimodal inputs with up to 1M context tokens.
OpenAI CSAM Image Classifier is now available at no cost for approved organizations
OpenAI made its CSAM Image Classifier available at no cost to approved organizations. The tool returns separate classification signals for potential child sexual abuse material and potential AI-generated CSAM to assist human review processes.
Budget limit increased to 40,000 budgets per account
GitHub increased the maximum budget limit for GitHub Enterprise Cloud accounts from 10,000 to 40,000 budgets. The limit applies automatically on github.com and GHE.com to support organizations managing spending across many users and cost centers.
Workforce Identity Federation is GA for Looker Google Cloud core instances
Workforce Identity Federation is now generally available as the primary authentication method for Looker (Google Cloud core) instances using public secure, private, or hybrid connections. It can be configured using the Google Cloud console or gcloud CLI during instance creation or updates.
GitHub Copilot models deprecated on November 16, 2026
GitHub deprecated Claude Opus 4.8, Claude Opus 5, and GPT-5.6 Terra across most GitHub Copilot features. Alternative models including Claude Opus 5.5 and GPT-6 variants are automatically enabled for Copilot Enterprise and Business subscriptions under standard settings.
More from Together AI platformEvery Together AI platform release →
Default expiry for Jig secrets
New Jig secrets now expire after 14 days unless you set a shorter lifetime with --ttl . Before a secret expires, the project's admins and the secret's creator get a daily email reminder. An expired secret is deleted, so new replicas can't read it. Existing replicas keep the value until they terminate.
New serverless models
Qwen/Qwen4-Max is now available on serverless. It supports a 1,000,000 context length, reasoning, function calling, and structured outputs.
New models available for fine-tuning
NVIDIA-Nemotron-3.5-Lightning-30B-A3B and NVIDIA-Nemotron-3.5-Super-VL-120B-A12B are now available for fine-tuning. Both models support function-calling and reasoning.
What's new:
Together AI has updated its model resource management to force cleanup of resources when context managers exit. Additionally, evaluation commands have been removed from the CLI following the service shutdown.
What's new:
- Tinker 0.39.0 compatibility: The Tinker-compatible client now builds on tinker 0.39.0, which the tinker extra installs. It follows Tinker 0.39's call styles, so inside a running event loop you can await the same calls you make in synchronous code. ServiceClient now works as a context manager that closes every training client it created.
New serverless models
Together AI launched tencent/Hy4 on its serverless platform, providing 1M context support, NVFP4 quantization, reasoning capabilities, multimodal input, and structured outputs.
Post-training: Nemotron 3.5 Super VL
Together AI added post-training support for NVIDIA Nemotron 3.5 Super VL. The platform supports LoRA training on B200-SXM GPUs with context windows up to 163,840 tokens, rank up to 32, and up to three adapter sessions per resource.
New dedicated endpoint models
The following models are now available for deployment on dedicated endpoints : Qwen/Qwen3-TTS-12Hz-1.7B (text-to-speech). This model was previously listed as Qwen/Qwen3-TTS-12Hz-1.7B-CustomVoice . thinkingmachines/Inkling-2-Small .