Together AI

Together AI cut Post-Training pricing on B200 GPUs to $9 per GPU-hour for the first eight weeks, down from $11.

@togethercompute on X

Together AI added DeepSeek V4.1 Flash 1017, an update tuned for building complete projects and agentic coding, at the same price as before.

@togethercompute on X

Together AI launched Together Post-Training, which lets developers turn open models into custom models using SFT or RL while Together runs the infrastructure.

@togethercompute on X

Together AI added Reflection's Beam, an open-weight coding and agentic model with 501B parameters and a 1M-token context window, under Apache 2.0.

@togethercompute on X

Together AI added NVIDIA Nemotron 3.5 Super VL, an open-weight multimodal model for visual reasoning across documents, screens and video.

@togethercompute on X

Together AI added Meta's Muse Spark Open, an open-weights model for agentic coding, to its serverless inference on launch day.

@togethercompute on X

Together AI launched Together Link, which runs open models in coding harnesses, tracks spending and uses an Auto router to pick the model.

@togethercompute on X

Together AI is offering 40% off Qwen3.7-Max and Qwen3.8-Flash on serverless inference through September 30.

@togethercompute on X

Together AI released tev1-4B-experimental, a classifier fine-tuned on Qwen3.5 4B, on its serverless platform at $0.042 per million input tokens.

@togethercompute on X

Together AI added canary, blue-green and rolling rollouts to Dedicated Model Inference, letting users upgrade a live endpoint's model in gated steps without downtime.

@togethercompute on X

Together AI added MiniMax H3, a 33B omni-modal video model that generates 4 to 15 second clips at up to 2K with stereo audio.

@togethercompute on X

Together AI added DeepSeek V4.1 Flash, a 552B-parameter model with a 1M context window and native multimodal input.

@togethercompute on X

Together AI fine-tuning now supports GLM-5.3 and Kimi K2.7-Code, adds live run metrics and early stopping, and cuts prices 30 to 70% on selected models.

@togethercompute on X

Together AI launched preemptible compute for its GPU clusters in public preview, at 50% of the on-demand price with up to 5 minutes to checkpoint.

@togethercompute on X

Together AI's open-model platform is now live across Equinix's global data centers as Equinix Inference Exchange, built with Equinix and NVIDIA.

@togethercompute on X

Together AI cut the price of Dedicated Inference on H100s from $5.49 to $3.99 per hour for September, applied automatically to existing deployments.

@togethercompute on X

Together AI added GLM-5.3 from Z.ai on launch day, with stronger coding and long-horizon agent performance.

@togethercompute on X

Together AI added GLM-5.3 Flash, Z.ai's first natively multimodal GLM-5 model, with 320B parameters, 18B active and a 1M context window.

@togethercompute on X

Together AI now offers Qwen3.8 27B for fine-tuning and Dedicated Model Inference, so users can train it on their own data and deploy it on dedicated infrastructure.

@togethercompute on X