Together AI launches Post-Training platform for agentic RL with tool-use environments
togethercompute · x · 2026-10-31
Together AI launched Together Post-Training, a managed service for RL and SFT on open models, with a focus on agentic RL — training across long, multi-turn, tool-calling interactions.
Key details:
- You define environments, prompts, rewards, verifiers and judges; Together handles GPU capacity, training–rollout consistency, parallelism, and tuned rollout engines
- Supports GRPO, PPO, or custom losses; multi-LoRA or full-weight training; async rollouts; bring-your-own sandbox
- RL loop abstracted as sample() → yourreward() → forwardbackward() → optimstep() → weightssync(); SFT reuses the same loop minus sampling and weight sync
- One call reserves all GPUs; parallel LoRA experiment sessions on a single reservation; W&B-compatible tracking
- Tinker-compatible, with cookbook recipes, multi-LoRA eval, and dedicated endpoints
More from Infra
- Google beams Earth prompt to TPU on orbiting satellite, proving space AI data centers work — ns123abc · 2026-10-31
- Google Cloud spend caps go GA with API, live cost reports and automatic monthly resets — rseroter · 2026-10-31
- Swift 1.5 Flash Next Heretic quants add IQ4_XS and IQ3_S tuned for reverse engineering — jjusko20 · 2026-10-31
- ClickHouse discloses unauthenticated RCE vulnerability CVE-2026-116366, patches available — Hacubu · 2026-10-31
- Count document passes, not tokens: batching questions cuts LLM scoring cost — petrusenko_max · 2026-10-31
- Budget dual-RTX 3060 12GB local inference rig beats a 5060 16GB setup on VRAM per dollar — Frizzy-MacDrizzle · 2026-10-31