Releasesby GlyphDeprecated llama-4-maverick-17b-128e-instruct See Deprecations for more details
Deprecated llama-4-maverick-17b-128e-instruct See Deprecations for more details.
Also shipped Oct 15, 2025Cerebras in October 2025 →
POS UI extensions: Developer experience Improvements
We’ve released a major update to POS UI extensions, aimed at reducing friction and speeding up your workflow. Key improvements include: Hot reloading improvements: We've made hot reloading faster and smoother. Code changes appear on your POS test device instantly without flickering.
Container Images (release 251015)
- quay.io/deepgram/self-hosted-api:release-251015 Equivalent image to: quay.io/deepgram/self-hosted-api:1.166.0 quay.io/deepgram/self-hosted-engine:release-251015 Equivalent image to: quay.io/deepgram/self-hosted-engine:3.104.7 Minimum required NVIDIA driver version: >=570.172.08 quay.io/deepgram/self-hosted-license-proxy:release-251015 Equivalent image to...
This Release Contains The Following Changes
- Adds support for Flux , our new turn-based streaming speech-to-text model! Flux is purpose-built for voice agent use-cases. Because of its resource usage, Flux must be run in its own self-hosted deployment separate from any other Deepgram models, even other speech-to-text models like Nova-3.
Push Make project to GitHub
You can now push your Make project directly to a new GitHub repository.
Tailscale GitHub Action v4.0.2
- Fixed: The Tailscale GitHub Action no longer logs the output of all shell commands to the runner's console unless debug logging is enabled. Fixed: The Tailscale GitHub Action masks authentication secrets in logs unless debug logging is enabled.
The product variant limit is now 2048 for all merchants
It is now possible for all Shopify merchants to create products with up to 2,048 variants in Shopify, exceeding our historical variant limit of 100. Please note that merchants using apps that are not using the in-support GraphQL product APIs may have a downgraded or broken experience when creating or viewing products with more than 100 variants.
More from Cerebras InferenceEvery Cerebras Inference release →
Added support for retrieving log probabilities via the logprobs parameter for the Completions endpoint
Added support for retrieving log probabilities via the logprobs parameter for the Completions endpoint. Set to an integer (0-20) to return the most likely alternative tokens at each position, or null to disable.
We're gradually introducing multi-token streaming across our models
We're gradually introducing multi-token streaming across our models. Instead of sending tokens individually, we'll now deliver them in batches through 200 evenly-spaced events per second. This change eliminates the artificial delays that occurred with single-token streaming.
Updated support for OpenAI GPT-OSS (gpt-oss-120b) - Tool calling with strict: true (constrained decoding) -...
Updated support for OpenAI GPT-OSS (gpt-oss-120b) Tool calling with strict: true (constrained decoding) Response format with json_schema when using strict=true Response format with json_object when using strict=true tool_choice: none Updated JSON Schema limitations: Nested levels increases from 5 to 10 Max properties increases from 100 to 500 Enabled...
Deprecated qwen-3-coder-480b
Deprecated qwen-3-coder-480b. We recommend migrating to Z.ai GLM 4.7.
We now support Predicted Outputs for faster generation when large portions of the output are known in advance
We now support Predicted Outputs for faster generation when large portions of the output are known in advance. Provide a draft of the expected response via the prediction field, and the model will efficiently reuse unchanged tokens while regenerating only those that differ. Ideal for code editing, document revisions, and template-based responses.