October 15, 2026
- Added Nemotron 3.5 Super VL to the API for training and inference. Read the announcement.
October 11, 2026
- Launched River Inference: call DeepSeek V4.1 Flash and GLM 5.3 Flash through an OpenAI-compatible API with your River API key, 20–60% cheaper than other providers. See the inference quickstart and models and pricing.
October 7, 2026
- Training metrics and projects in Console: log loss, rewards, evals, or any other number from your training script, and compare runs side by side in a project. See the metrics guide.
September 3, 2026
- Added GLM 5.3 Flash to the API.
August 21, 2026
- Added DeepSeek V4 Flash and Nemotron 3.5 Lightning 30B to the API.
July 6, 2026
- GLM 5.2 launched (serving + training configs); Teams support with shared-wallet credit management.
June 29, 2026
- Dropless MoE training support for GLM, Kimi, and Qwen model families.
June 22, 2026
- GLM 5.1 correctness fixes and speculative decoding; retired-model filtering in chat.