The Deploy Log | Week by week | Free, no signup
273 rows are listed for in The Deploy Log, from 65 companies. OpenAI shipped the most, 28 rows, then Amazon Web Services with 25. Models first seen in the record this week: Claude Haiku 5.5, DeepSeek V4.1 Flash, MiniMax M3 and Mistral Large 4. Every item sourced to the publisher's own page.
| Company | Rows |
|---|---|
| OpenAI | 28 |
| Amazon Web Services | 25 |
| GitHub | 22 |
| Anthropic | 20 |
| Databricks | 15 |
| Vercel | 13 |
| Google Cloud | 8 |
| LangChain | 8 |
| Cloudflare | 7 |
| Microsoft | 7 |
| Figure | Value |
|---|---|
| Rows | 273 |
| Rows the week before | 310 |
| Change against the week before | -37 |
| Companies | 65 |
| Companies seen for the first time | 17: Amazon, AMD, Charm, Diffusers, Docker, Figma, FlashAttention, Gradio, IBM, Midjourney, MLflow, Promptfoo, Red Hat, SambaNova, Samsung, SAP and Twilio |
| Models seen for the first time | 4: Claude Haiku 5.5, DeepSeek V4.1 Flash, MiniMax M3 and Mistral Large 4 |
| Rows with no shipped date, placed by their edition or the day they were added | 20 |
| Lane | Rows | Rows the week before |
|---|---|---|
| Models and APIs | 26 | 31 |
| Tools and products | 238 | 270 |
| Robotics, hardware and chips | 6 | 3 |
| Breakthroughs | 0 | 2 |
| Funding and acquisitions | 0 | 2 |
GPT-6 with Intelligent UI rolls out globally to ChatGPT Plus, Pro, Business and Enterprise tiers today, expanding to Free and Go tiers tomorrow. The model composes responses with graphics, buttons, forms, charts and interactive experiences, and GPT-6 Instant starts answering web search questions 44% sooner than GPT-5.6 Instant.
Claude Haiku 5.5 is available today on Amazon Bedrock through US, EU, AU, JP and Global inference profiles, and costs around 75 percent less than Claude Haiku 4.5 for most tasks. It is the first Haiku model with effort controls for tuning cost against intelligence per task.
Cloudflare released Clef-omni, an open-weight decision model that takes audio in wav or mp3, video in mp4 or webm, image and text input. Cloudflare also cut the price of Clef-flash so it is now cheaper than Jev, and made Clef faster.
Google expanded SynthID Detector globally in English, letting anyone check if an image, video or audio was made with AI from Google or partners including OpenAI, NVIDIA, Kakao and soon Apple, after watermarking over 180 billion images and videos and 240,000 years of audio.
Microsoft AI developed a local version of MAI Code 1.1 Flash, a 137 billion total and 6.8 billion active parameter mixture-of-experts model, with the quantized on-device version coming in at 53GB, an 80% reduction in size, and scoring 70.80% on SWE-Bench Verified.
Sophos cut average response time for cases using agents from approximately 38 minutes to 89 seconds, a 96% reduction, and now resolves 52% of MDR cases end-to-end with AI built through OpenAI Daybreak.
Asana used GPT-6 Astra in Codex to optimize StackAI browser agent on GPT-6.1 Sol, cutting estimated model cost 76x to $0.47 per run and making runs 5x faster at roughly four minutes, with 89% of input from cache at 5% of the uncached price.
NVIDIA released a GR00T N1.7 SSD placement deployment model with three cameras on Hugging Face.
NVIDIA released the agile_one_s_walk_n2_1680 walking deployment model checkpoint on Hugging Face.
Qwen-Image-2.1-Turbo, an accelerated 7B text-to-image and image editing model with 8 denoising steps, is available on Hugging Face.
Anthropic launched the Cyber Mission with the Critical Infrastructure Defense Program and free OSS Scanner for open-source projects.
Mistral released a 51-class Arabic language and dialect classifier on Hugging Face with download and Python usage instructions.
Thinking mode reruns 8.2 jobs to fix missed prompt details, and folders can now be shared with collaborators or viewers.
StepFun Step 5 Preview is available on OpenRouter at $1/$2.70 per 1M tokens with 1M context.
Mistral released a 4.4B parameter streaming Arabic speech-to-text model with 8.82% average CER at 480 ms delay.
Claude Haiku 5.5 is live on OpenRouter at $0.10/$0.50 per 1M tokens with a 1M context window.
Adds lyria-3.5, gemini-omni-1.1-flash, and gemini-omni-flash-preview to Interactions Model enum, plus continuation_token support.
Improved Nova-3 Medical English models for batch and streaming with fewer word errors and fewer errors on medical entities.
Spaces in Google Flow Music lets creators build custom instruments or effects.
Google Nano Banana 2.1 image generation and editing model is available on OpenRouter at $1.50/$30 per 1M tokens.
Mistral Large 4 is available on OpenRouter with a 512K context window and 256K output tokens.
Nova-3 Medical now supports multilingual transcription across 10 languages for batch and streaming.
GLM 5.3 is now available on Amazon Bedrock.
Improved Nova-3 monolingual models for Catalan, French, Indonesian, and Vietnamese are live for all users.
Untied copy of Qwen3-0.6B-Reverse-Text-SFT with tie_word_embeddings false for trainers that don't support tied LM heads.
Untied copy of Qwen3-0.6B with tie_word_embeddings false for trainers that don't support tied LM heads.
Untied copy of Qwen3-4B-Instruct-2507 with tie_word_embeddings false, loadable by trainers that don't support tied LM heads.
Untied copy of Qwen3.5-4B with tie_word_embeddings false, loadable by trainers that don't support tied LM heads.
Perplexity released pplx-decider-v1.1-27b on Hugging Face, raising Decision Index from 56.4 to 61.56.
46 new datasets are available on the Registry of Open Data on AWS.
LiteLLM v1.105.0 released with 747 commits to main since the previous release.
Qwen Code Desktop v0.25.1 adds Qwen Live endpoint settings, conversation search, and MCP image bounds.
RunPod ships an official MCP server for AI IDEs including Claude Code, Cursor, and VS Code.
openai-python v3.29.0 adds a billing_not_active session error code, shida Live voice, and vault updates and metadata filters.
Microsoft Decision-1 is now available on AI Gateway, answering typed questions about text with probabilities, choices and scores for classification, workflow routing or rubric assessment, with no separate Azure account or deployment required.
Liquid d1 is now available on AI Gateway, a decision model that evaluates shared state against typed questions for classification, routing and scoring, returns structured answers with probabilities without generating text tokens, and supports vision for image classification and inspection.
Grok Imagine Video 1.5 Lite from xAI is now available on AI Gateway, turning text prompts and images into video with native audio in clips from 1 to 15 seconds at 480p, 720p and 1080p across seven aspect ratios.
AI Gateway now supports OpenAI's Ultrafast service tier for GPT 6 Astra and GPT 6.1 Sol, billed at 6x the standard per-token rate, with requests pinned to unsupported regions such as the EU running at the standard tier.
FLUX 3 Image from Black Forest Labs is now available on AI Gateway, generating images from text or editing existing images with up to ten reference images and five resolution tiers from 768 by 768 to 4K, with no Black Forest Labs account required.
Step 5 Preview from StepFun is now available on AI Gateway, built for agentic coding, professional knowledge work and financial analysis, with text and image input and a 1M-token context window.
GitHub Copilot added Claude Haiku 5.5 for Pro, Pro+, Max, Business and Enterprise users, made local sandboxing generally available in Copilot CLI, the Copilot app and VS Code sessions using Agent Host, and let the Copilot app use separate GitHub accounts for license and repositories.
CodeQL 2.27.2 is now available, adding a C++ regular-expression parser, SQL-injection sink models for the Comdb2 C API, Go websocket modeling, Rust async data flow improvements, and JavaScript Workflow SDK directive recognition, with the Default suite running 498 security queries covering 170 CWEs.
Claude Code v2.1.296 added a code key to the Claude apps gateway managed policies, an autoCompactWindow option for subagents, a CLAUDE_CODE_WORKFLOW_SUBAGENT_MODEL environment variable, and updated Sonnet 5.5 cache reads to $0.10 per million tokens from $0.20.
Codex rust-v0.162.1 fixed a TUI crash when asynchronous questions contain multiple lines, preserving line breaks and complete hyperlink destinations, and fixed startup failures from differences between a running background server's feature settings and CLI defaults.
Ollama v0.40.2 upgrades models downloaded with earlier versions in the background the first time you run them for better performance and compatibility on llama.cpp, keeps the original copy as a backup, and makes ollama list stop showing duplicate entries for upgraded models.
Google made Carrier Link available in 14 additional countries including the US, Canada and much of Europe, letting Workspace customers add phone numbers and calling plans from certified local carriers through partners Southlight Services, Gamma Communications, Tata Communications and Telefonica Global Solutions.
Google Workspace Studio added nested conditionals, Gmail skip replies for thread triggers, Google Meet calendar outputs with full invitee lists and post-meeting email draft templates, and a Search for a Doc Drive step, rolling out gradually starting .
Amp now supports using your Claude Pro/Max subscription in Claude Code mode, free for everyone.
Model IDs in /model and /config are now case-insensitive and canonical IDs are saved.
Interactive sandbox settings now suggest possible environment secrets and let you add masking hosts before saving.
Open WebUI v0.12.0 adds realtime voice calls using OpenAI Realtime voice model gpt-realtime-2.1-mini.
The OpenAI Java SDK adds agent environment webhook events and vault updates.
RunPod's new REST API lets users manage GPU workloads programmatically, including launching, scaling, and monitoring pods.
llama.cpp release b11551 adds MiniCPM-V 4.7 model support.
Alteryx Live Query now integrates with BigQuery, enabling Gemini-powered extraction from PDFs and images with warehouse-native execution.
Amazon Bedrock now supports reasoning summaries for OpenAI models.
Amazon Connect Customer now provides automated checks to improve performance evaluation forms.
Amazon EC2 R8g instances are now available in additional regions.
Amazon EC2 R8gd instances are now available in additional regions.
Go SDK v1.80.0 adds types for Chat and Cowork unified analytics metrics and Managed Agents workflows, multiagent configuration, and thread status filtering.
Java SDK v2.71.0 adds types for Chat and Cowork unified analytics metrics, Managed Agents workflows, multiagent configuration, thread status filtering, and client options().
Available in asia-northeast1 (Tokyo) and supports gemini-3.5-flash.
Claude Agent SDK TypeScript v0.3.296 adds claude_code_version to initialize responses and autoCompactWindow to AgentDefinition.
Claude Haiku 5.5 is now available on DigitalOcean Inference for serverless inference.
Anthropic Claude Sonnet 5.5 and Claude Opus 5.5 are now available on Kiro in AWS GovCloud (US).
Clef-omni adds audio and video input, Clef-flash is cheaper, and Clef is faster.
On-demand CPU and memory profiling with flamegraphs for Workers and Durable Objects.
Fixed copilot config supports sandbox credential injectHosts keys, with key completion in Bash, Zsh, and Fish.
Crush v0.98.1 routes newer Copilot models through the Responses API.
Docker agent v1.150.0 adds DaoXE provider alias, OSC 7501 status reporting, execution capabilities, and a datetime toolset.
Dynamo v1.5.0 adds custom Router scoring, Mooncake and disk KV indexing, and guided tool call streaming.
Genie One can now run code in a secure, isolated sandbox for work beyond SQL.
GitHub Copilot CLI 1.0.95-1 adds native Microsoft Entra broker authentication on macOS with browser fallback.
Pre-release v1.0.96-0 improves interactive session startup, permission decision visibility, and model failure recovery.
Managed Google Workspace connector is available in Beta in Databricks Lakeflow Connect for ingesting audit activity.
LangSmith SDK v0.14.7 released with sandbox API endpoint prefix normalization.
Langfuse v4.56.0 adds syntax highlighting to session tool previews and spend alerts for ClickHouse-billed organizations.
Mistral v2.26.1 adds session read-only status, proxy setup, narrator, and plugin commands.
ONNX Runtime CUDA Plugin EP 0.3.0 adds sparse attention, packed continuous-batching, INT2 and mixed-width QMoE execution, and decode-kernel tuning.
Word, Excel, and PowerPoint creation and collaboration are now available inside the Microsoft Copilot app.
OpenAI Java SDK v4.79.0 adds prewarmed hosted environments API support.
Databricks lets users select OpenJev (Qwen3.5 4B) as the evaluator model service for custom LLM-as-a-judge service policies.
Pydantic AI v2.55.0 requires Python 3.11 or newer and adds durable execution for ExaSearch, YouSearch, and LocalStack tool requests.
JavaScript SDKs now support Anthropic, OpenAI, Google GenAI, Vercel AI, LangChain, and LangGraph with automatic instrumentation since 11.0.0.
TwelveLabs Pegasus 1.5 model is now available on Amazon Bedrock.
Custom service policies in Unity Gateway can now check caller-supplied request tags before a model request or MCP tool call.
Workspace Studio adds nested conditionals, Gmail skip replies, Meet calendar outputs, and Drive document search, rolling out from .
Types for Chat and Cowork unified analytics metrics, plus workflows, multiagent configuration, and thread status filtering for Managed Agents.
Native Microsoft Entra broker authentication on macOS, sandbox credential injectHosts keys, and --context for ACP sessions.
ogx v1.5.0 adds Text-Embeddings-Inference provider, DeepSeek and Fireworks AI passthrough, and Anthropic reasoning model thinking support.
openai-node v7.31.0 adds prewarmed hosted environments.
API adds agent environment suspension and expiration.
openai-python v3.27.0 adds prewarmed hosted environments.
openai-python v3.28.0 adds agent environment suspension and expiration.
openai-ruby v0.102.0 adds prewarmed hosted environments.
openai-ruby v0.103.0 adds agent environment suspension and expiration.
Fixes hidden Accordion revealed with open=True rendering empty.
Adds an app view to gr.Workflow and reuses up-to-date upstream results when running a node.
Default automatic MCP tool discovery now caps at 64 pages, with explicit maxListPages for larger listings.
Fixed Vercel authentication configuration for restored sessions.
Fixed audio interruption after generation with transport-appropriate buffer clearing.
Patch release updates dependencies to @openai/agents-core@0.20.0, agents-realtime@0.20.0, and agents-openai@0.20.0.
AWS Cost Explorer, Budgets, and Dashboards now support Amazon Bedrock product attributes.
Agno v3.1.2 adds conversation compaction to keep long sessions inside the context window.
GuardDuty RDS Protection now detects data exfiltration and destruction in Aurora and RDS databases.
Azure Database for PostgreSQL flexible server is generally available in East US 3.
CLI v3.0.70 fixes 400 errors on GPT-6 Astra, GPT-6.1 Sol, Claude Fable 5, and Claude Opus 5.5, and fixes MCP server exit timing.
Python SDK v0.2.165 on PyPI, bundling Claude CLI 2.1.294.
Claude Agent SDK TypeScript v0.3.294 updates to parity with Claude Code v2.1.294.
Claude Code v2.1.294 fixes prompt and agent hooks written as instructions and improves Stop/SubagentStop hook judgment.
Added onFailure block for hooks, Program Status Protocol support, and quoted text in /copy picker.
Anthropic Claude Opus 5.5 and Claude Sonnet 5.5 are available in Gemini Enterprise AI developer tools.
Claude Haiku 5.5 added to model selection and --model completions.
Crush v0.98.0 adds x.AI OAuth login, OSC 7501 support, and many fixes.
GPU device plugins are no longer enabled by default on clusters with GPU nodes; DRA drivers are opt-in.
The Discover page and domains are now generally available on Databricks.
DeepSeek V4.1 Flash Fast is available on Baseten Model APIs with 1M-token context and dedicated Fast capacity.
Prompt caching for DigitalOcean-hosted models is generally available with automatic context caching and no cache-write charge.
Disconnected cluster analysis is available now with pen drive powered by Red Hat Lightspeed.
Adds LLM-planned DAG workflows with Durable Functions execution, retries, and crash recovery.
Experimental snapshot build serving moonshotai/Kimi-K3 from an SGLang runtime on Dynamo 1.5.1 and SGLang v0.5.18.
FastMCP v4.1.0 released with fixes across OpenAPI, proxies, auth, tasks, CLI, and full Python 3.15 support.
File upload in Genie One is now generally available for CSV, Excel, PDF, image, Word, and plaintext files.
OpenAI GPT-6.1 Sol now supports Ultrafast mode on Amazon Bedrock.
Gemini assistance in the BigQuery Studio SQL editor is generally available through an inline action button.
File upload in Genie One is generally available for CSV, Excel, PDF, image, Word, and plaintext files.
Added Claude Haiku 5.5 to model selection and --model completions.
Bugfix release restoring typed tool HTTP header validation.
Adds an app view to gr.Workflow and reuses up-to-date upstream results when running a node.
JetBrains extension adds prompt navigator, MCP OAuth sign-in, and markdown agent board messages.
smithtune CLI turns LangSmith trajectories into fine-tuned models via Fireworks or Baseten.
Langfuse v4.55.0 adds OpenAI decision model support and skill import from local files and ZIPs.
Adds user-owned credentials, user-level memory, HTTP channels, Slack file transfer, and Parallel web search.
Managed StandardV2 NAT Gateway for AKS is generally available.
Expanded quantized inference and attention support, kernel performance improvements, and reliability fixes.
OpenHands v1.26.0 released with new features including verify-openhands skill and backend workspace explanations.
AWS launched an open-source stack for building physical AI machines, with architecture guidance, deployment automation, and code.
Microsoft Power Pages is generally available in Sweden Central and Italy North.
Promptfoo 0.124.1 adds OpenAI Decisions support and exposes catalog text-token pricing.
Retool 4.34.8 and Retool 4.0.21 are now available on the Stable release channel.
Fused LM head on by default, up to 6.9x longer sequences on the same GPU.
Added skipMiddlewareRequestBody option to reduce Fast Origin Transfer usage and improve time to first byte.
harness-cli v0.2.0 adds a vended subagent tool and exposes a tool client accessor.
harness-python v0.2.0 adds a vended subagent tool and exposes a tool client accessor.
Python SDK v1.59.0 adds a vended subagent tool and fixes Anthropic cache writes on older LiteLLM.
TypeScript SDK v1.20.0 adds a vended subagent tool and exposes a tool client accessor.
harness-typescript v0.2.0 adds a vended subagent tool and exposes a tool client accessor.
huggingface_hub v2.2.0 adds hf jobs stats snapshots, concurrent xet downloads, and HfFileSystem caching fixes.
langchain-openai 1.7.0 adds support for the decisions API.
Adds BigQuery authorized views, LookML dashboards, Spanner Omni, and a looker-get-explore tool.
OpenAI Node SDK v7.30.1 corrects custom voice creation parameters.
supervision 0.30.9 installs without OpenCV, fixes COCO/CreateML box loading, and improves video frame reading.
Updated unsafe topic detection categories are now available in cf.llm.prompt.unsafe_topic_categories.
Stabilizes Anthropic agent package and adds fixes for session state, search memory, and request messages.
C8gb instances are now generally available in additional regions.
Go SDK adds claude-haiku-5-5 and typed computer and browser toolset tool calls.
Java SDK adds claude-haiku-5-5 and typed computer and browser toolset tool calls.
Anthropic TypeScript SDK v0.132.0 adds claude-haiku-5-5 and typed computer and browser toolset tool calls.
BigQuery now offers TimesFM 3.0 as a built-in time series forecasting model with AI.FORECAST, including multivariate forecasting.
Added optional subagent_type to background_tasks_changed task entries and updated to parity with Claude Code v2.1.293.
Claude Code v2.1.293 adds Claude Haiku 5.5, now the default Haiku model, with 1M context and $0.10/$0.50 per Mtok.
Claude Haiku 5.5 is live in Devin, scoring 58.4% on FrontierCode 1.1 at about an eighth of Sonnet 5's cost per task.
Claude Haiku 5.5 is generally available in GitHub Copilot for Pro, Pro+, Max, Business, and Enterprise users.
Claude Haiku 5.5 is now available on Vercel AI Gateway with effort levels and Zero Data Retention support.
Claude Haiku 5.5 is now available on AWS.
Claude Haiku 5.5 is now available on AWS GovCloud (US).
Claude Haiku 5.5 is available through Netlify AI Gateway and Agent Runners with zero configuration required.
Anthropic Claude Haiku 5.5 is now available on Unity Gateway via Foundation Model APIs.
Cline v4.1.23 adds GPT-6.1 Sol to recommended models and changes default models for 20 providers.
Cloudflare One Client for Windows 2026.8.2100.0 is GA with faster reconnections, lower memory use, and service recovery.
Cloudflare One Client for macOS 2026.8.2100.0 is GA with faster connects, lower memory use, and split tunnel fixes.
Cloudflare Organizations is generally available for Enterprise customers and MSSP/Distributor partners, with nested sub-organizations for partners.
Command sandboxing is available to all users via /sandbox and --sandbox.
Databricks Apps now GA with on-behalf-of-user authorization for permission-aware apps.
Deep Agents can bind tools to skills, pin skills at runtime, and reload skills mid-thread.
DeepSeek V4.1 Flash now supports priority pay-per-token and reserved provisioned throughput on Databricks.
Managed databases and Storage-Optimized Kafka/OpenSearch plans are available in Memphis (mem1).
Adds GitHub repository skill loading, evaluator routing through the models gateway, and model pricing and capability fixes.
Patch release fixes router overload recovery, min_tokens, Qwen3-VL video routing, and RL worker discovery.
Pre-release fa4-v4.0.0.beta34 with Windows CUDA 13 and PyTorch 2.15 build fixes.
The microVM sandbox type is generally available with GKE Sandbox in clusters running version 1.37.0-gke.4713000 and later.
GitHub Copilot CLI 1.0.93 adds enterprise permissions.limitTo, safe command execution during active turns, and localhost allowlist changes.
GitHub Copilot CLI version 1.0.94-0 adds /model to discover supported models from a running local Ollama instance.
MCP server configuration changes apply between turns without restarting the session.
GitHub MCP Server 2.0.1 fixes the Go module path for library users.
Glyph Cluster reasoning model is available in stealth for free on AI Gateway for Pro and Enterprise plans with purchased AI Gateway credits.
Kilo Code v7.8.8 adds pinned worktrees, in-app web links, and reduced Agent Manager header height.
Released v4.54.0 with trace classification, external media storage, and session transcript views.
Local sandboxing for GitHub Copilot is generally available in CLI, app, and VS Code.
Workspace admins and owners on paid plans can now set credit usage limits and alerts for workspaces, projects, members, groups, or access tokens.
MLflow 3.17.0 adds typed judges, fine-grained permissions, and faster analytics for large trace datasets.
Managed Deep Agents v0.9 adds agent-created schedules, per-run configuration, and Slack reactions.
OpenAI Decisions API now available through Netlify AI Gateway, running on gpt-6-luna with zero configuration required.
Vercel AI Gateway adds an OpenAI-compatible /v1/decisions endpoint for GPT-6 Luna Decisions and other decision models.
Retool 4.70 is now available on the Edge release channel for self-hosted organizations.
SAP Commerce Cloud, cloud ERP edition is generally available, with new and enhanced Joule Assistants for SAP Customer Experience.
SAP is rolling out Joule Work to all customers this month, a new UI for AI.
Stdio MCP servers get a 10s default initialize budget, up from 3s.
Retool 4.34.7 is now available on the Stable release channel for self-hosted instances.
Vercel Flags now supports timestamp attributes for entities, configurable via dashboard or CLI.
Adds claude-haiku-5-5 and typed computer and browser toolset tool calls.
Pre-release 1.0.94-0 adds update guidance for managed settings and lets managed policy disable Assisted Permissions.
Adds --image-uri to deploy an already-pushed image and a 'langgraph deploy listeners list' command.
AWS Certificate Manager now supports ACME issuance through AWS PrivateLink.
Canvas authoring plugin for Copilot is available in the Awesome Copilot marketplace to generate Azure canvas projects.
Updated to parity with Claude Code v2.1.291; install via npm, yarn, pnpm, or bun.
Claude Agent SDK TypeScript v0.3.292 adds agent_id, parent_task_id, run_id, and typed sections to ListAgents.
Claude Code v2.1.292 adds marketplace install, Agent effort parameter, overloaded retry delay, and prompt.autocomplete.
Diffusers 0.41.0 adds Qwen-Image 2.1 pipeline with text-to-image, image editing, native transparency, and LoRA training.
DigitalOcean Harness Runtime now supports CrewAI agents in an isolated sandbox.
The Figma agent is generally available with improved latency, file search, and configurable guidelines.
Gemini Nano Banana 2.1 is generally available, optimized for high-speed multimodal image generation and editing at 1K, 2K, and 4K output resolutions.
Google's Gemini Nano Banana 2.1 is now available through Netlify AI Gateway, callable from Netlify Functions without configuring API keys.
Genie Code CLI, a terminal coding agent for Databricks data and AI work, is in Beta.
GitHub MCP Server 2.0.0 adds structured outputs and output schemas for clients advertising MCP spec support or later.
Point-in-time editing is now available in Google Vids, making the canvas match the video timeline 1:1.
Patch release restores legacy task execution on SDK 2.x client and preserves reported cache token usage.
Langfuse v4.52.0 bounds trace batches by estimated weight and shows organization usage breakdown on every deployment.
AI gateway now exposes resolution context in response headers, spans and logs.
Lightwell Clearinghouse is generally available, letting businesses request priority review and remediation for open source software.
Users can comment on specific lines in the diff on the Code tab of Agent Runners.
Mistral Large 4 is available on AI Gateway via one API key, with AI SDK, OpenAI-compatible, Responses, and Anthropic Messages APIs.
Mistral Vibe v2.26.0 adds built-in plugin discovery, image attachment warnings, background process panel, and owner-only .env fallback.
Google's Nano Banana 2.1 image model is available on Vercel AI Gateway via AI SDK, OpenAI-compatible API, or AI CLI.
OpenAI Go SDK v3.72.0 adds agent turn items and usage source grouping.
OpenAI Node SDK v7.29.0 adds local tool failure observation, session stream tool running, and tool search support.
Users can now run multiple separate Puck conversations, each with its own Puck, switching between them.
Patch release fixes two cases of silently wrong training and three crashes, including end-of-turn handling for GRPO, RLOO, and Distillation.
Trello for Google Workspace now lets users create Trello cards from Google Chat messages.
Pre-release 1.0.93-2 adds enterprise permissions.limitTo and prioritizes GPT-6.1 Sol, GPT-6 Astra/Luna, and Claude 5.5 models.
livekit-agents 1.8.5 adds Gemini 3.8 Live support and sticky LLM fallback routing.
llama.cpp b11454 adds K2 Horizon dense and MoVA model support.
openai-go v3.73.0 adds standalone Decisions support.
openai-java v4.78.0 adds standalone Decisions support.
openai-node v7.30.0 adds standalone Decisions API support.
openai-python v3.26.0 adds standalone Decisions API support.
openai-ruby v0.100.0 adds agent tool search and turn items, and selects Mantle for GPT-6.1 Sol.
supervision 0.30.8 fixes video quality, YOLO label parsing, mask drift, and VLM parsing.
vLLM v0.31.1rc0 exposes cached prompt tokens by cache tier in metrics.
MCP TypeScript SDK client package version 2.3.1 released.
MCP TypeScript SDK codemod package version 2.3.1 released.
MCP TypeScript SDK core package version 2.3.1 released.
MCP TypeScript SDK server package version 2.3.1 released.
Adds optional rollback item budget for Responses session compaction and fixes trace exports and generation reasoning.
Fixes conditional tool approvals, durable approval resumes, uncopyable normalized values, and default function redaction.
AWS Capabilities by Region now offers availability notifications for individual features and advanced filters.
Amazon EC2 Hpc8a instances are now available in Asia Pacific (Singapore).
Iceberg external and managed tables now support flexible column names by default.
Claude Code v2.1.290 adds serverToolUses, agentId, ceiling, and ThemeKey/Color types to plugin hooks.
ComfyUI v0.39.0 adds LynnReal light Minimax-H3 VAE support, DynamicGroup widget input, and higher quality defaults for Save Video encoding.
Databricks SQL version 2026.39 is rolling out to the Current channel.
Hook-driven agent routing, lean TUI settings panel, codemode tool call visibility, and new CLI and eval capabilities.
Docker agent v1.148.0 fixes oversized tool output handling by bounding complete tool results to 50 KiB and fixes duplicate line rendering in the lean TUI.
Genie Code scheduled tasks are now available by default for workspaces with the compliance security profile enabled.
Copilot CLI 1.0.92 adds config subcommands, a pre-conversation environment picker, and fixes for MCP connections and voice runtime errors.
Lakeflow Jobs can now use up to five triggers, with combined types and independent pause.
Lakeflow pipelines can now set run-as identity to an account group, in addition to user or service principal.
Langfuse v4.51.0 adds decision-model evaluators, comparison grid annotations, and AI gateway request ID correlation.
Workspace admins on Business and Enterprise plans can now show an announcement banner to every member of the workspace.
requireBearerAuth in server-legacy now accepts the optional expectedResource parameter.
OpenAI Agents JS v0.17.1 adds server-wide MCP tool guardrails, custom output guardrail blocked messages, Docker sandbox labels, and image results in web search.
Qwen Code Desktop v0.25.0 released with session diagnostics, managed runtime attestation client, trajectory time range selection, Qwen Live endpoint settings, and private/group access policies.
SDK TypeScript v0.1.18 released, bundling CLI version 0.25.0.
Amazon Redshift now supports creating and refreshing Apache Iceberg materialized views.
Segment Conversation Memory Integration is now in general availability.
langgraph 1.2.13 released with fixes for checkpoint replay, update_state, and DeltaChannel counters.
llama.cpp v0.6.0 adds GLM-5.3-Flash 320B, Clef, Qwen4Exp MTP, and a new /v1/systemone server API.
The Frame hospitality model is now available in the U.S. in 55-, 65-, and 75-inch sizes.
Galaxy Tab S12 Ultra and Tab S12+ are available starting today, starting at $1,399.99 and $1,199.99.
Prompt caching for MiniMax M3 on SambaCloud, cached tokens billed at $0.06 per million tokens, 90% below standard input rate.
MuJoCo 3.15.0 released with .mjz archive path fixes and actuator dampratio computation changes.
ROCm 10.1 adds asynchronous fast-path storage, NUMA-aware host memory allocation, and ROCm CLI with AMD Skills.
SmartThings update adds SmartThings Now cards, redesigned Home tab, and expanded local control.
One line every time AI actually ships, around the clock, free on Threads and Bluesky. Every one is in The Deploy Log with its source.