DeepSeek V4.1 Flash 1017 Ultrafast is now available in public preview on Vercel AI Gateway. The model supports a 1M-token context window and is optimized for high-speed interactive coding and agentic workflows.
Lovable apps can now use GPT-6.1 Terra for reasoning and image analysis, and GPT Image 2.6 Flare and Sunburst for image generation and editing.
Gemini Omni 1.2 Flash is now generally available for video generation, replacing Veo 3.1. The update improves video clarity and instruction following.
OpenRouter released the Auto Router and Auto Code router. These tools evaluate factors like live pricing, provider health, and benchmarks to select an appropriate model based on governance policies.
Moonshot AI's Kimi K3.1 Pro is now available on Amazon Bedrock. The model features a 1-million-token context window and supports prompt caching for agentic tasks like coding and document processing.
Vercel added DeepSeek Search and DeepSeek Fetch tools to AI Gateway, enabling web search and Markdown content extraction across tool-calling models.
Braintrust now supports in-browser thread translation and updates its CLI to support optional MCP telemetry and improved trace continuation. The update also adds support for recording model variants in coding-agent sessions.
Gemini Omni 1.2 Flash is now generally available, offering omni-modal processing for text, image, and video. New features include standalone video upscaling, extended generation durations, and reference-guided video editing.
Gemini Omni 1.2 Flash is now available on AI Gateway, enabling text-to-video generation and editing through a unified API.
Cloudflare is deprecating several AI models on Workers AI, including select DeepSeek, Llama, Gemma, and Qwen variants, with complete removal scheduled for November 15, 2026.
OpenAI released GPT-6 Cyber, a model specialized in identifying exploitable vulnerabilities and developing security patches while operating within authorized scopes.
Starting January 2027, AI Functions in SQL queries will require Databricks Runtime 17.3 LTS or higher. This change does not affect ai_similarity or the AI Functions REST API.
Databricks retired the DeepSeek V4 Flash (0731) model. Users are advised to migrate to DeepSeek V4.1 Flash (1017).
Web Answer is now generally available on Amazon Bedrock AgentCore. It provides a managed tool that returns synthesized, cited answers grounded in web sources in a single call.
Mistral AI introduced Mistral Studio and Le Chonk to enable building and deploying enterprise agents, workflows, and applications with control over models, data, and infrastructure.
Perplexity added an ultra effort level option to its Agent API, designed for broad research and data-gathering tasks using Claude Opus 5.5, with support for background execution mode.
Together AI added the moonshotai/Kimi-K3.1-Pro model to its serverless platform, supporting a 1M token context window, function calling, and structured outputs.
Databricks announced that starting January 2027, AI Functions requiring Databricks Runtime 15.4 LTS or higher (excluding ai_similarity) will mandate Databricks Runtime 17.3 LTS or above.
Databricks jobs now support AI Function tasks for parsing, extracting, or classifying files and table rows without writing SQL, currently in beta.
GPT Image 2.6 Sunburst and Flare models are now available in Netlify AI Gateway, providing automatic authentication and caching for Netlify Functions.
The Gemini Enterprise mobile app now supports switching between multiple teams without re-authentication, and allows renaming or removing teams from the list.
- Changelog for v0.65.0 chore(release): bump version to 0.67.0-nightly.20261020.gb041de017 Changelog for v0.66.0-preview.0 feat(cli): bridge the IDE companion into container sandboxes over exec stdio fix(core): prevent duplicate agent loading when workspace is home directory fix(ide): exclude IDE settings and GEMINI.md files from open file con text...
Cohere made Embed 5 (Pro and Fast) and Rerank 5 (Pro and Fast) models available on OpenRouter, supporting multimodal inputs and unified embedding spaces.
Cloudflare AI Gateway introduced model aliases, enabling applications to direct requests to configurable model pools with fallback and smart routing.
Daybreak Red, featuring the GPT-5.6 Cyber model, is now available to authorized customers in AWS GovCloud (US-East) via Amazon Bedrock for cybersecurity workloads.
OpenAI released GPT-Image-2.6 Sunburst and Flare models, supporting adjustable complexity, SVG generation, animated GIF creation, and markup-guided editing.
The Kimi K3.1 Pro model is now available on DigitalOcean Inference for both serverless and dedicated inference.
Dify 1.19.0 brings nested workflow tracing to your observability tools. Follow a workflow into the Workflow Tools it calls, inspect the steps inside each invocation, and keep the LLM context alongside the execution tree.
Vercel AI Gateway expanded Zero Data Retention provider support for Moonshot AI Kimi K3.1 Pro and K3.1 Pro Highspeed models, supporting 1M token contexts and multimodal input.
Together AI now supports LoRA post-training for the Kimi K3.1 Pro model with a 163,840-token training context.
The GitHub Copilot usage API now distinguishes between Copilot-hosted models, local BYOK usage, and enterprise custom models, providing detailed metrics at the user and aggregate levels.
Pinecone document indexes now support Google's gemini-embedding-2 and OpenAI's text-embedding-3-small and text-embedding-3-large models.
Langfuse 4.75.0 increases AI Gateway limits to 10k requests per minute, simplifies Codex setup, and adds analytics for gateway members.
AIP added experimental support for DeepSeek V4.1 Flash (1017) through Fireworks on non-georestricted commercial enrollments. The open-weight multimodal model supports text and image inputs, tool calling, and a 1-million-token context window with US-hosted inference.
Hy4 from Tencent is now available on AI Gateway , at 50% off through December 31, 2026. Hy4 supports reasoning , tool calling , structured outputs , and image input, and has a 1M-token context window for work across a codebase or many documents. Hy4 is the first Hy model with vision. It can pick out fine detail in an image and read complex charts.
The gemini-4-argon model is now generally available in AI Functions.
Admins can now connect Claude to the Mixpanel MCP Server for every member at once. When members open Claude, Mixpanel is already connected, with no sign-in or consent screen. With this update: One-time setup: A Claude admin and a Mixpanel organization admin connect the two. In Claude, use the connector's Managed authorization settings.
With Claude auto-connect, members with a matching PagerDuty account are connected to PagerDuty in Claude automatically, with no sign-in screen and no identity provider (IdP) configuration. Members act as themselves in PagerDuty, with the permissions they already have there.
Connecting a Salesforce data source with Gemini Enterprise using data federation is generally available (GA). You can use the Gemini Enterprise app assistant to search across your Salesforce accounts, contacts, opportunities, cases, leads, and custom objects, get answers grounded in that content, and create, update, and delete records without needing to...
Text generated by Claude Mythos 5, Claude Opus 4.7, and Claude Sonnet 4.6 carries Anthropic’s text watermark on the platforms listed in How Claude marks AI-generated content. On Claude Sonnet 4.6, text generated for a request that sets temperature, top_p, or top_k to a non-default value might not carry it. See Text watermark.
CodeQL is the static analysis engine behind GitHub code scanning, which finds and remediates security issues in your code. We’ve recently released CodeQL 2.27.3 . It supports new Python 3.15 language features, adds JSON syntax for data extensions, adds C23 bit-precise integers, and includes accuracy improvements across languages.
Model Vault and Model Vault Encrypted now run leading third-party open models alongside Cohere's models, on fully managed, dedicated infrastructure. You can choose the right model for each task and run every one of them with the same protections.
Lovable now follows the AGENTS.md and CLAUDE.md files in your project's subfolders . When Lovable opens or edits a file, it reads the instruction files in that file's folder and in every folder above it.
Users can now evaluate AI agents and compare model performance within the Datadog Agent Observability Playground using integrated Fireworks AI models.
You can now tune SQL queries using AI. Available starting with version 1.6.2 of the google_ml_integration extension, the ai.tune_sql function uses an LLM along with query execution plans, schemas, and runtime stats to recommend query optimizations.
Serving tuned Gemini 3.x models on global endpoints is available in General Availability (GA) . You can copy fine-tuned Gemini 3.x models to the global region using the Model Registry models:copy API, deploy them to global serving endpoints, and send inference requests directly to aiplatform.googleapis.com .
Long conversations and agent loops on the Messages API can now send smaller requests and get a lower time to first token with message threads, in beta with the message-threads-2026-08-12 beta header. The API keeps the conversation on the server, so each later request sends only the new messages.
Datadog is now available on the Claude Marketplace, providing visibility into Claude-powered AI applications.
LongCat 2.5 from Meituan is now available on AI Gateway. This model combines native visual understanding with coding and multi-step tool use for agents working across complex codebases and full-project builds.
Any MCP client can now ask to connect to Vercel MCP . You still log in and approve each connection, but the client no longer has to be reviewed by Vercel first. Previously, Vercel reviewed and allowlisted each client and its redirect URIs by hand, and anything not on the list was rejected.
A new Anthropic model is now available in the Kiro IDE and CLI for the AWS GovCloud (US) Regions. With Haiku 5.5, all three Claude 5.5 models are now available in Kiro in AWS GovCloud (US). Claude Haiku 5.5 is the fastest model in the Claude 5.5 family.
Atlas Agent Engine is now available in public preview. It provides a secure way to run AI agents at scale.
InclusionAI's Ling 3.1 Flash 1019 model is now available on AI Gateway, featuring a 1M-token context window for coding and agent workflows.
The OpenAI GPT-6.1 Terra model is now available for serverless inference on DigitalOcean.
Fast mode for Claude Opus 4.8 is deprecated and will be removed on November 30, 2026. Users should migrate to Claude Opus 5.5 to continue using fast mode.
Cortex Training is now available in preview, allowing users to customize open-weight LLMs using a serverless service within Snowflake.
GPT-6 Rosalind is a specialized model for biological reasoning and scientific tool use. It is available to approved organizations for internal research.
OpenAI is implementing expanded violence prevention safeguards and earlier interventions in ChatGPT to improve safety for teen users.
Daybreak Red, a cyber defense model from OpenAI, is now available on Amazon Bedrock in AWS GovCloud (US-East). It is designed for authorized vulnerability research and exploit reproduction.
Neon has added support for Google's Gemini 4 Argon model to its AI Gateway, featuring a 1M token context window.
Releasesby Glyph