Data and AI releases

2,002 this month · 180 companies

What's new:

BetaNot yet verified

- 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.

ObservabilityDeveloper toolsRegions

Read on docs.together.ai

Also shipped Nov 18, 2026Together AI in November 2026 →

More from Together AI platformEvery Together AI platform release →

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.

GA
Source ↗

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.

GA
Source ↗

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.

Update
Source ↗

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.

Update
Source ↗

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.

Update
Source ↗

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.

Update
Source ↗

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 .

Update
Source ↗