Releasesby GlyphLegacy API key regeneration
Legacy API keys can now be regenerated by any admin or editor of the default project. Legacy keys appear in the project's API keys table with a Deprecated badge. See Legacy API keys .
Also shipped Aug 27, 2026Together AI in August 2026 →
You can now train models using XGBoost version 2.1 by using the XGBOOST_VERSION option
You can now train models using XGBoost version 2.1 by using the XGBOOST_VERSION option. BigQuery uses the XGBoost 3.1.0 library to load and make predictions on XGBoost models. This feature is generally available .
CoCo in the Snowflake Extension for Visual Studio Code (General availability)
CoCo in the Snowflake Extension for Visual Studio Code is now generally available. CoCo runs in a side panel next to the extension's existing SQL and Snowpark Python workflows, so you can work locally while you stay connected to your Snowflake account.
Lakeflow Designer August Release
The Lakeflow Designer August release adds the Unique operator, expands the Aggregate, Select, and Python operators, and improves the canvas, preview, and version-history experience.
Genie Code is now available as a Lakeflow Jobs task (Beta)
You can now run prompts with Genie Code as a task in a Lakeflow Job. Genie Code runs your prompt autonomously. It can read upstream task outputs and call tools as needed. The task returns a link to the resulting Genie Code conversation with its response. See Genie Code task for jobs .
Genie Code for AI Runtime (Public Preview)
Use Genie Code to develop and troubleshoot deep learning workloads in notebooks connected to AI Runtime. Now in Public Preview , it generates distributed training code, resolves environment and dependency issues, and debugs GPU and distributed workload failures. See Use Genie Code with AI Runtime .
Share foreign Delta tables with OpenSharing (Public Preview)
You can now use OpenSharing to share foreign Delta tables, which are Delta-format tables registered in a foreign catalog such as OneLake, a Hive metastore, or AWS Glue. Providers add foreign Delta tables to a share, and recipients access the shared data in read-only format.
More from Together AI platformEvery Together AI platform release →
Billing usage API (beta)
GET /billing/usage returns organization-wide usage and cost line items, filterable by month. The endpoint is in beta and enabled per organization. See Get billing usage .
File upload progress in the CLI and Python SDK
tg files upload and tg files check now show a progress bar in interactive terminals. The Python SDK accepts an optional progress_callback on client.files.upload() that receives upload progress events. See the files CLI reference and Data preparation .
Slurm cluster kubeconfig for project members
Project members can now download the cluster kubeconfig for Slurm clusters from the cluster details page, matching Kubernetes cluster behavior. See Download cluster kubeconfig .
Model deprecations
The following models have been deprecated and are no longer available on serverless: nvidia/Nemotron-3-ultra-550b-a55b . pearl-ai/gemma-4-31b-it . deepseek-ai/DeepSeek-V4-Pro . Use deepseek-ai/DeepSeek-V4-Pro-0813 instead. moonshotai/Kimi-K2.7-Code .
New serverless models
The following models are now available on serverless : zai-org/GLM-5.3 : 1,000,000 context length, FP4 quantization, function calling and structured outputs. Pricing: $1.40 input / $4.40 output / $0.26 cached input (per 1M tokens).
Batch jobs in the CLI
The Together CLI now includes a tg batches command group for batch inference : tg batches submit uploads a local JSONL file (or takes an existing file ID) and creates a job against chat.completions , audio.transcriptions , or audio.translations . tg batches list , retrieve , and cancel manage the job lifecycle, with ls and get as aliases.
New serverless models
The following models are now available on serverless : zai-org/GLM-5.3-Flash : 1,000,000 context length, FP8 quantization, function calling and structured outputs. Pricing: $0.15 input / $0.50 output / $0.03 cached input (per 1M tokens).
active_sessions autoscaling metric for TTS
Dedicated endpoint autoscaling now accepts active_sessions , which targets concurrent open client WebSocket sessions per replica on a TTS deployment ( AVERAGE_VALUE only). The default inflight_requests metric also covers TTS handlers, so HTTP TTS traffic can keep scaling without a policy change. See Configure autoscaling .