'%3e%3crect%20x='16'%20y='51'%20width='68'%20height='27'%20rx='13.5'/%3e%3ccircle%20cx='32'%20cy='53'%20r='15'/%3e%3ccircle%20cx='53'%20cy='43'%20r='20'/%3e%3ccircle%20cx='71'%20cy='57'%20r='12'/%3e%3c/g%3e%3c/svg%3e){kind=small}
Cloud Changelog
Together AI platform changelog 2025
27 announcements, 2025
All Together AI platform changes · 271 announcements since July 2025 · 1 GA
Free. Unsubscribe any time.
RSS feed for Together AI platform
December 20254
November 20252
October 20252
September 20258
Batch inference adds UI and all serverless models, with 30B enqueued token limit
Fine-tuning context lengths increased for DeepSeek-R1-Distill, Gemma 3, Qwen and Llama models
Fine-tuning adds gpt-oss-120b, DeepSeek-V3.1 and Llama 4 models
Fine-tuning supports any CausalLM under 100B parameters from Hugging Face Hub
Self-service NVIDIA GPU clusters with API-first provisioning, including B200 and H100
Evaluations now run on serverless LoRA models and dedicated endpoints
Kimi-K2-Instruct-0905, an upgraded 1 trillion parameter MoE model, added
August 20254
July 20257
VirtueGuard-Text-Lite guard model added, with 8ms response time
Evaluations benchmarking platform added, using LLM-as-a-judge
Qwen3-Coder-480B-A35B-Instruct-FP8 added, with a 256K context length
DeepSeek-R1-0528 serverless inference faster on NVIDIA Blackwell, same pricing
Moonshot AI's Kimi-K2-Instruct 1T-parameter MoE model now available
Whisper Large v3 audio transcription 15x faster than OpenAI, files over 1 GB
Together AI achieves security compliance through independent audit