The Deploy Log | Week by week | Free, no signup
246 rows are listed for in The Deploy Log, from 73 companies. OpenAI shipped the most, 25 rows, then Amazon Web Services with 21. Models first seen in the record this week: GPT-6.1. Every item sourced to the publisher's own page.
| Company | Rows |
|---|---|
| OpenAI | 25 |
| Amazon Web Services | 21 |
| Vercel | 19 |
| GitHub | 14 |
| Google Cloud | 12 |
| LangChain | 11 |
| Microsoft | 9 |
| Anthropic | 8 |
| Databricks | 7 |
| Qwen | 7 |
| Figure | Value |
|---|---|
| Rows | 246 |
| Rows the week before | 319 |
| Change against the week before | -73 |
| Companies | 73 |
| Companies seen for the first time | 11: Alphabet, Arm, Baidu, Celestica, Dropbox, Lightricks, Salesforce, Sierra, Triton, Vertiv and Weaviate |
| Models seen for the first time | 1: GPT-6.1 |
| Rows with no shipped date, placed by their edition or the day they were added | 7 |
| Lane | Rows | Rows the week before |
|---|---|---|
| Models and APIs | 19 | 26 |
| Tools and products | 217 | 276 |
| Robotics, hardware and chips | 2 | 10 |
| Breakthroughs | 0 | 1 |
| Funding and acquisitions | 5 | 4 |
Claude Code v2.1.316 added pooled group budgets to the Claude apps gateway, CLAUDE_CODE_SUBAGENT_EFFORT, and guardrail.input_tagging for Bedrock upstreams. It deprecated fast mode for Opus 4.8 as of , with removal on November 30, 2026.
Google Fitbit Edge is on shelves now with the largest, brightest tracker display and a new screen design, physical side buttons, and a wrist-contoured shape with accessories.
Mistral released the weights of Mistral Large 4, a 1-trillion-parameter multimodal model, after three weeks of additional reinforcement learning. It leads open-weight coding on FrontierCode 2.0 with 19.7% and Terminal-Bench 4.0 with 41%, and scores 51% on the Artificial Analysis Cyber Index.
The ChatGPT desktop app is now available on Amazon Bedrock in AWS GovCloud (US), routing Codex model inference through Amazon Bedrock Mantle in us-gov-west-1 with IAM governance and zero data retention support.
Mistral Large 4, a 1T-parameter MoE model with 49B active parameters and 1M context, is on Hugging Face.
Adds continuation_token, video upscale, and video understanding.
Adds continuation_token, video upscale, and video understanding.
Adds continuation_token, video upscale, video understanding, and Python 3.15 support.
PixelSeal and AudioSeal marking weights and configs for LTX-2 are available on Hugging Face.
Open-weight model for multilingual transcription, speaker attribution, translation, and audio Q&A, now available on Hugging Face and ModelScope.
Particle-1 is available on OpenRouter at $0.25 per 1M input tokens and $1 per 1M output tokens with 1.0M context.
Released an 8.13M parameter streaming guardrail probe for GLM-5.3-Flash with under 0.5% decode overhead.
Ling-3.1-flash-1019 released with 560B total parameters, 25B active, 1M context.
ONNX inference exports of the unified Metatron G1 scene/recovery policy, with demo and racer files.
Released a 5.18M parameter streaming guardrail probe for Qwen3.8-Flash-Next with under 0.5% decode overhead.
Adds 25 languages including 11 Indic locales and claims up to 2x transcription speed.
GPT-6.1 Terra is available with standard API prices 50% lower than GPT-5.6 Terra and GPT-6.1 Sol.
GPT-6.1 Terra is available on OpenRouter at $1/$5 per 1M tokens with 1.1M context.
OpenAI GPT-6.1 Terra Pro is available on OpenRouter at $1/$5 per 1M tokens with 1.1M context.
Liquid AI released LFM3-1.7B-Base, a 1.73B parameter pre-trained text model with 256K context, on Hugging Face.
OpenBMB released MiniCPM-OCR-1.0-DSpark, a 296M parameter draft head that speeds up MiniCPM-OCR 1.0 decoding by 55%.
Unlimited-OCR-2 released on Hugging Face with SR-SWA for unlimited-horizon document parsing.
CrewAI 1.15.34 adds standalone LiteLLM catalog, crewai skill search, deploy create --zip, and multiple bug fixes.
KoboldCpp 1.124.1 adds YuE2 music gen and BreezeTTS 2, Qwen3-TTS, VoxCPM2, and Ming 3.0 TTS models.
Added support for Responses WebSocket warmup.
Fixed connection_pool version incompatibility with Ruby 3.3.0.
Bounded SQL history queries, Generated UI on by default, Web Shell keyboard shortcuts, Russian locale, and one-click daemon updates.
Bounded SQL history queries, Generated UI on by default, Web Shell keyboard shortcuts, Russian locale, and one-click daemon updates.
Mistral Large 4 on Vercel AI Gateway now uses updated weights with lower pricing for input, cached input, and output tokens, and the existing model ID string mistral/mistral-large-4 serves the new version without changes.
GitHub added three webhook delivery IP ranges to the Meta API hooks field: 9.234.98.160/28, 9.234.98.176/28, and 9.234.106.48/28, for github.com across all plans including GitHub Enterprise Cloud.
Codex rust-v0.166.0 added keyboard-driven transcript selection with word, line, and page motions, integrated search, GPT-6.1 Terra in bundled, Vertex AI, and Amazon Bedrock model catalogs, and codex queue --resume.
Vercel IP geolocation now resolves city-level data for nearly all requests and postal-code data for 90% of traffic, with more frequent refreshes, enabled for all projects with no code or configuration changes.
Vercel Flags SDK initialization on cold starts dropped from over 70 ms to 10 ms at p75, roughly 7x faster, and usage reporting no longer keeps functions running after the response, with update @flags-sdk/vercel to 1.5.3 or later.
Added permissions.blockReadsExemptPaths to the Settings type and updated to parity with Claude Code v2.1.316.
KoboldCpp 1.124 adds YuE2 music generation and BreezeTTS 2, Ming 3.0, Qwen3-TTS, and VoxCPM2 TTS models.
OpenAI Node SDK v7.47.0 adds typed Responses WebSocket warmup support.
Gemma 4 images are resized without letterbox padding in llama.cpp release b12021.
Added agent context compaction, session review, time awareness, image edit overlays, and SIP DTMF commands.
Adds agent context compaction, session review, time awareness, hosted tool usage, and more.
Adds agent context compaction, session review, time awareness, image edit overlays, and more.
Added agent context compaction, session review, time awareness, image edit overlays, and SIP DTMF commands.
Adds agent context compaction, session review, time awareness, image edit overlays, and more.
playwright-mcp v0.0.84 adds isolated browser contexts per tab and redacts runtime URL parameters.
Ling 3.1 Flash 1019 from InclusionAI is available on Vercel AI Gateway free until , with a 1M-token context window for coding, multi-step reasoning, and tool-using agents.
Released dotnet-1.27.0 with a .NET consumer skill and updated AgentServer dependencies.
Browser Run now runs Chrome 152 by default with updated User-Agent, better site compatibility, and other fixes.
claude-agent-sdk 0.2.175 bundles Claude CLI 2.1.314 and is available on PyPI.
Added sandbox_required_unavailable, permissions.blockWritesOutsideWorkingDirectories, and prewarm().
Claude Code auto mode is now supported for both Anthropic and non-Anthropic models hosted by Databricks.
Claude Code v2.1.314 adds permissions.blockWritesOutsideWorkingDirectories, gateway upstream groups, Bedrock guardrail_trace, and idle compaction for 200K+ token conversations.
Daybreak Red from OpenAI is now available to eligible customers on Amazon Bedrock in AWS GovCloud (US-East).
GPT-6.1 Terra is now available on DigitalOcean Inference for serverless inference.
Gemini 4 Argon is now available on the AI Gateway with a 1M-token context window and text, image, video, audio, and PDF input.
Gemini in Google Slides can now edit multiple slides or an entire presentation at once with a single prompt.
Spend caps on Google Cloud Budgets are generally available with API automation, live cost reports, and automated monthly resets.
Web Search is now available on Kiro in AWS GovCloud (US).
New Copilot Home navigation, projects, Workspaces, Autopilot, role-based plugins, and more rolling out to Frontier in October and worldwide in November.
Release v3.0.4 adds /release-notes, preserve_api_base, VIBE_TERMINAL_KEY, credential source display, system prompt file replacement, and output token caps.
Amazon SageMaker AI real-time endpoints now support GPU partitioning.
TensorRT Edge-LLM CLI runs Nemotron-3.5 Lightning on DGX Spark 64GB, with DSpark speculative decoding reaching 126.9 tokens per second.
CPU ONNX Runtime and OpenVINO prediction is up to 2x faster on some systems.
AutoBatch now sizes from the train dataset for every task, with fixes for model loading, results saving, datasets, and deployment workflows.
Prebuilt Qwen CUA Driver binaries for macOS, Linux, Windows, and Node.js SDK packages.
Python SDK v1.63.0 adds Bedrock maxTokens default, Model.model_info, structured output attempt limits, and a graph tool.
TypeScript SDK v1.23.0 adds Bedrock maxTokens default, Model.model_info, structured output attempt limits, and a graph tool.
Harness TypeScript v0.3.3 adds Bedrock maxTokens default, Model.model_info, and a summarizer fallback fix.
langgraph 1.2.19 released with streaming fixes and ToolNode tool names in graph node metadata.
langgraph-checkpoint-sqlite 3.1.3 released with a fix rejecting NUL bytes in namespace labels.
langgraph-sdk 0.4.11 released with stream reconnect retries and SSE state preservation fixes.
Ollama v0.40.13 brings Liquid AI LFM3 models, including text-to-speech in lfm3-audio and lfm3-omni, to Linux and Windows in addition to Apple silicon.
Node.js 26 LTS is generally available for builds and Vercel Functions, with HTTP/2 by default for built-in fetch via Undici 8, the Temporal API, and V8 14.6, and it is the default for all new projects.
langchain-core 1.7.2 added add_notification_metadata and get_notification_metadata and fixed merge_content so it no longer changes the first message's content.
Native approvals for Harness Agents now reach the AgentOS approval UI for Claude and Codex.
Amp can now visualize or build charts, timelines, diagrams, and clickable mockups in the thread.
BigQuery can now query Databricks-managed Apache Iceberg tables via Lakehouse with metadata refresh as often as every 5 minutes.
CrewAI 1.15.33 adds skill search, deploy create --zip, and fixes for token usage and JSON crew loading.
Edge 150+ on Windows 11 25H2/26H2/26H1 with October 2026 update gets up to 33% faster launch.
GPT 6.1 Terra is now available in Retool through OpenAI for AI resource queries and building apps.
GitLab Orbit is generally available, with agents using up to 45x fewer retries and 4.5x fewer tokens in testing.
GitLab Orbit is generally available for querying vulnerability blast radius across projects, owners, and pipelines.
HfFileSystem now supports copying and moving files into and within Buckets.
Langfuse v4.72.0 adds AI Gateway analytics, PagerDuty alerts, trace-level eval filters, and AI Gateway onboarding.
Workspace admins and owners now get an email when credits are likely to run out before the monthly refill.
Managed Deep Agents v0.10 adds browser use and async subagents, available today.
Model Armor is available in Brazil (southamerica-east1) with prompt injection, responsible AI, and sensitive data protection filters.
Expanded attention, mixture-of-experts, and quantized operator support with GPU kernel performance improvements.
Patch release with model-validation, memory-safety, Plugin EP data transfer, and build fixes.
Open WebUI v0.12.5 adds analytics session explorer, multi-page notes, live presence in notes, and rename from chat menu.
Pipecat v1.15.0 adds NVIDIA TTS custom_configuration and classifier-based user turn stop strategies.
Qwen Code v0.26.4 adds unscoped npm extension installs, independent worktrees, and inline model timing.
Amazon RDS for Oracle and SQL Server now support storage autoscaling on additional storage volumes.
Real-Time Mode for Spark Declarative Pipelines is generally available on Databricks.
SAP Business Data Cloud now supports native Apache Iceberg and a universal catalog powered by Apache Polaris.
Amazon SageMaker AI now supports SOCI-indexed images to start training jobs in seconds.
AWS Step Functions workflows are now available as AI agent tools.
Strix can now run on any of the 400+ models on Vercel AI Gateway.
TRL v1.17.0 adds LoRA fine-tuning support for Mistral Large 4 on a single 8×B300 node.
Transformers v5.22.0 adds support for Mistral Large 4, a 1T-parameter MoE model with 49B active parameters.
Vercel Connect now supports webhook triggers for Resend, X, Twilio, Notion, and Telegram.
Vercel Firewall now lets you log or block AI crawlers and AI assistants separately via the AI Bots managed ruleset.
Vercel Plugin is now available in the Google Antigravity marketplace with skills for AI SDK, Vercel Functions, and Workflow.
Adds structured output attempt limits, agent.cancel() stop, and graph, read, write, edit tools.
Adds structured output attempt limits, agent.cancel() stop, and graph, read, write, edit tools.
Adds structured output attempt limits, agent.cancel() stop, and graph, read, write, edit tools.
Released @gradio/utils 0.15.0 with full screen layout support for gr.workflow.
Released @gradio/workflowcanvas 0.14.0 with ZeroGPU token forwarding and workflow layout fixes.
AWS Agent Registry is now available in four additional AWS Regions.
Amazon Bedrock AgentCore CLI version 1.0 is now available.
Three agent skills for Application Integration are available in the public google/skills repository.
GitHub Copilot Fusion can now route between supported on-device models and cloud models in CLI, app, and VS Code.
Copilot UX components are generally available with SharePoint Framework 1.24, enabling interactive dashboards, forms, and business actions in Copilot.
GitHub Copilot Fusion, local sandboxes, and next edit suggestions are now available in JetBrains IDEs.
The Dataverse plugin is now available in Copilot Cowork, building governed Dataverse apps from one prompt.
Reachability analysis for Dependabot is in public preview for JavaScript and TypeScript, adding call-path verdicts to alerts.
Docker agent v1.154.0 adds Lunaroute as an OpenAI-compatible provider with credential-based selection and model discovery.
New Echo Dot with built-in smart home hub and Echo Studio with Home Camera are available now.
Workspace admins and owners on Business and Enterprise plans can now extend their workspace with Events & Hooks.
Gemini Enterprise app is available in the Brazil (br) region with at-rest data residency and in-region ML processing, GA with allowlist.
GitHub Agentic Workflows is now generally available for automating reasoning-based tasks with coding agents inside GitHub Actions.
Computer use is in public preview on macOS and Windows, letting Copilot read app content, click controls, and navigate across apps.
Cloud agent sessions, Agent Host, and cross-client session continuity are now available in GitHub Copilot for VS Code.
The GitHub security extension is in public preview, letting users find, explain, and fix code scanning alerts in the Copilot app.
Gradio 6.31.0 released with video format conversions and ZeroGPU X-IP-Token forwarding.
New MacBook Pro with PROLED XDR Display, M5 Pro/M5 Max, and Touch is available to pre-order today, with availability beginning .
Particle-1 from Particle AI is now available on AI Gateway with a 1M-token context window.
Paseo now supports running fx, Codex, Claude Code, OpenCode, and Pi in the cloud with the Vercel Sandbox plugin.
Prime 360 cinematic highlights run on Amazon Trainium chips during select NBA on Prime broadcasts this season.
Qwen Code Desktop v0.26.3 adds session permission rules, auto approval, an experimental shared agent loop, and CLI fixes.
Raycast v2.10 adds images to Notes and lifts the 5 note limit on the free plan.
Batch inference is generally available on RunPod Serverless with up to 3 million requests a day and up to 30% off on-demand pricing.
Four migration and modernization assistants are available to early adopter customers for cloud ERP transformation projects.
Shared Copilot Automations are rolling out to Copilot Pro, Pro+, Max, Business, and Enterprise plans.
Shop SDK lead capture now accepts an authorization Details attribute to request specific profile fields, including phone number and birthdate.
Shopify Analytics now shows events from supported apps, including marketing emails and loyalty redemptions.
Shopify Analytics released bubble maps that group nearby shipping or billing addresses into bubbles sized by a chosen metric.
Campaign Autopilot is now available to all merchants with channels like Meta Ads, Microsoft Advertising, Snapchat, Shop, and Shopify Messaging.
Shopify Flow added a Product updated trigger that can monitor up to 5 product fields.
External agents can now create, update, delete, activate, and deactivate Shopify Flow workflows through AI Toolkit.
TorchVision 0.29.2 is a minor release with no new feature, compatible with PyTorch 2.14 and later.
v8.4.192 supports torch 2.15 CUDA in DDP training and ONNX Runtime inference.
Vercel Container Registry now supports up to 2 GB per compressed image layer, up from 500 MB.
Virtual Models let apps call a saved model configuration by ID and change the underlying model or settings centrally without redeploys.
Weaviate v1.40.3 fixes backup auth, telemetry timeouts, inverted index writes, shutdown cycles, and geo index queue deletion.
Selecting Copilot Fusion High or Max in /model now works from long-context sessions.
fx is now available as a coding agent provider in T3 Code on web, desktop, and mobile.
pydantic-ai v2.63.0 defaults CodeMode to dynamic_catalog=True and changes Gemini Live reconnect and Harness LLMReminder defaults.
qwen-code v0.26.3 improves tool call recovery, memory extraction, and context-window detection for qwen models.
supervision 0.31.3 adds grayscale annotator support and merges multi-part masks in YOLO export.
vLLM-Omni v0.32.0 adds Ray-based multi-node inference, session-aware replica placement, and multi-frontend full-duplex serving.
Metastore admins can use Unity Catalog permissions on the system.ai schema to control which AI Functions users can run.
Agent Evals, API Evals, and advanced cost controls for Agent Fabric, handling over 1.2 billion requests daily.
Agno v3.1.11 lets ClaudeAgent and CodexAgent be members of Teams and steps of Workflows.
Amazon Aurora Serverless now supports up to 384 ACUs.
Amazon ECS now supports GPU sharing with NVIDIA Multi-Process Service (MPS).
Higher limit of Azure Blob Storage accounts per region per subscription is generally available.
Claude Agent SDK v0.3.312 adds autoMode settings, unanswered annotations, and parity with Claude Code v2.1.312.
Cost management and tagging for Claude models in Microsoft Foundry is generally available.
Data Engineering Agent now accepts Google Sheets, CSV, or JSONL files from Google Drive as pipeline input.
ChatGPT, Codex, Claude, Claude Code, and Cursor can now connect to Databricks MCP servers using default OAuth client IDs.
Dropbox for Microsoft Copilot integration lets users find, open, summarize, organize, and compare Dropbox content in Copilot.
Data tiering is now available on R8gd and R7gd nodes in Europe (Spain) and Asia Pacific (Hyderabad).
Amazon ElastiCache for Valkey now supports changing durability on existing clusters.
OpenAI GPT-6.1 Terra is now available on Unity Gateway via Foundation Model APIs.
VPC connectivity and static IP egress for custom MCP servers, supported in us-central1 and europe.
Gemini Robotics Embodied Reasoning 2 is now available in public preview on the Gemini platform.
Copilot CLI 1.0.104 starts sessions on Auto, applies fusionTier settings, and enables Assisted Permissions without experimental mode.
GitHub Copilot supports server-to-server authentication for GitHub Apps on organizations in public preview, with read-only Copilot Requests permission and installation access tokens.
Workspace admins and owners on Business and Enterprise plans can now map IdP groups to workspace roles.
AI gateway now records model, key owner, byte and upstream error facts on generations.
Failure logs for rejected requests now keep only core request fields, tags and guardrail details.
Failure logs for rejected requests now keep only core request fields, tags and guardrail details.
Failure logs for rejected requests now keep only core request fields, tags and guardrail details.
NVIDIA RTX Spark Windows PC platform with up to 128 GB unified memory, DLSS 5, and Nemotron 3.5 Bolt E4B 4GB is available.
OCI Responses API now supports explicit and automatic context compaction for long-running conversations and agent workflows.
OpenAI Go SDK v3.89.0 adds prompt cache breakpoint TTL options.
OpenAI Node SDK v7.45.1 clarifies prompt cache breakpoint TTL behavior and supports google-auth-library v11.
Per-conversation Docker runtime and Settings → Runtime page added.
Pydantic AI v2.62.0 adds mid-conversation thinking effort changes and new model error categories.
Qwen Code v0.26.2 adds session permission rules, auto approval, and an experimental shared agent loop.
Remote MCP Server for Oracle workloads is now generally available, providing a unified interface for AI agents to run SQL on Oracle databases.
SageMaker MLflow Apps now support MLflow 3.16.
Triton 3.9.0 adds tl.atomic_load, tl.atomic_store, and tl.atomic_poll.
RKNN INT8 export now enabled for detection, segmentation, pose, and oriented bounding box models.
Unity Gateway Usage Dashboard v1.0 is now available with coding-agent adoption, harness usage, and estimated session spend analysis.
Vercel Observability now shows AI bot requests, crawled pages, assistant retrievals, markdown requests, and errors.
Vercel Plugin is now available in OpenCode 2 with skills for AI SDK, Vercel Functions, Workflow, and specialist agents.
Visual Studio 2026 Stable Channel adds commit-to-working-tree compare, bulk conflict resolution, and a new Copilot chat experience.
Workflows can now be restricted to eu, us, or fedramp jurisdictions.
langchain-anthropic 1.7.7 released with trace resolved reasoning effort fix and dependency updates.
openai-java v4.94.0 adds prompt cache breakpoint TTL options and native OkHttp client support.
Adds prompt cache breakpoint TTL options to the API.
openai-ruby v0.114.0 adds prompt cache breakpoint TTL options.
Pinecone released pg_pinecone, an Apache 2.0 Postgres extension for filtered vector search on existing pgvector columns.
Automated Snowpipe setup with Snowflake CoCo is now generally available, building ingestion pipelines from a single natural-language conversation.
Browser Use 0.13.14 adds OpenAI browser tools and a shared Bash helper.
CLI v3.0.74 attaches to existing runs, times out Hub health checks after 3s, and retries dropped turns at most 3 times.
Cisco IQ for Partners is now generally available with MCP server and expanded APIs.
Claude Code v2.1.311 adds --hold-idle-file and --max-hold-min to self-hosted-runner and fixes long freezes and context errors.
Desktop v0.0.49 adds a Changes button, project files panel, and integrated terminal.
Added Cloud tasks with a Local | Cloud toggle, cloud model selection, and sandbox progress card.
DNS settings endpoints now represent Advanced Nameservers with nameservers.type: "cloudflare.advanced", rolling out over seven days.
Devin Security 2.0 launches with frontier-model scanning, prioritization, and autonomous remediation that turns findings into pull requests.
FastMCP v4.1.4 fixes CLI transport preservation, session visibility, OAuth identity assertions, and adds redis-py credential provider support.
GPT-6.1 Terra is generally available in GitHub Copilot for Pro, Pro+, Max, Business, and Enterprise users.
OpenAI GPT-6.1 Terra is available on Vercel AI Gateway with a context window over 1M tokens and a fast mode.
OpenAI GPT-6.1 Terra is generally available on Amazon Bedrock.
OpenAI's GPT-6.1 Terra model is now available through Netlify's AI Gateway and Agent Runners with zero configuration required.
Assisted permissions are generally available in GitHub Copilot CLI, GitHub Copilot app, and Visual Studio Code.
IAM Helper Agent Skills and MCP Tools are now available to automate IAM workflows for agents.
VS Code 7.8.18 and CLI 7.8.18 with dependency updates and hosted cloud sessions exposed through local Kilo API.
Lakeflow Jobs now show job-level query performance metrics and insights for supported tasks.
LookML semantic language indexing in Knowledge Catalog for Looker Core is now generally available.
Skills added or edited on disk are picked up by running sessions without /reload.
Model Armor enforces the managed constraint modelarmor.managed.requireTemplateDataResidencyCompliance by default, now generally available.
Extends autonomous recovery engine to Kubernetes workloads managed by NVIDIA Run:ai.
OpenAI Java SDK 4.92.0 adds delegation controls and prompt cache lifetimes.
OpenAI Java SDK 4.93.0 adds the gpt-6.1-terra model identifier.
SAP reports 45,000 cloud customers using its AI solutions and expanded Joule Assistants and specialized agents.
SDK v0.2.0 reattaches to live sessions, records capability registrations, and retries dropped reasoning-only turns.
Sierra is now available on Google Cloud Marketplace.
Supabase Personal Access Tokens can now be scoped by resource, permission, and expiry.
YOLO26-seg training now uses Mask-IoU-aware label assignment, with fixes for export, inference, training utilities, and data handling.
Vercel Sandbox runCommand() now supports stdin with writeStdin() and closeStdin() in TypeScript and Python SDKs.
Custom monitors are available on Pro and Enterprise plans with Observability Plus at $0.0035 per GB scanned.
claude-agent-sdk-typescript 0.3.311 released with cache lifetime token fields and Node.js 18.14.0 minimum.
DeltaChannel now takes a full snapshot every 100 updates instead of 1000, with snapshot_frequency=1000 to keep old behavior.
langgraph-checkpoint 4.3.1 released, rejecting NUL bytes in namespace labels.
langgraph-cli 0.4.39 adds --json to deploy list.
CLI now sends the upload headers the control plane returns.
Retries runs.wait with a bodyless GET, follows SSE spec, and accepts durability, stream_resumable and context in runs overloads.
The OpenAI Go SDK adds delegation controls and prompt cache lifetimes.
The OpenAI Go SDK adds the gpt-6.1-terra model identifier.
openai-node 7.44.0 released, adding delegation controls, prompt cache lifetimes, and scheduled automations.
openai-node 7.45.0 released, adding gpt-6.1-terra model identifier.
Adds delegation controls, prompt cache lifetimes, and expands Live Decision item and question types.
Adds gpt-6.1-terra model identifier to the OpenAI Python SDK.
Adds delegation controls and prompt cache lifetimes to the openai-ruby SDK.
Adds gpt-6.1-terra model identifier to the openai-ruby SDK.
sherpa-onnx v2.0.1 adds VoxLock target-speaker voice isolation to speech denoisers.
Neural Dawn is now available to download and play on five mobile devices powered by Arm Mali G2-Ultra NX GPU.
NVIDIA Vera Rubin NVL72 and NVIDIA Vera CPUs are now generally available on Oracle Cloud Infrastructure.
AWS net sales rose 39% year-over-year to $46.0 billion, a $184 billion annualized run rate.
Google Cloud revenues increased 103% to $30.7 billion, led by AI infrastructure and solutions.
Oracle set FY30 total revenue guidance of $225 billion or more, over 35% CAGR from FY26, with non-GAAP operating margin of 40%+ and non-GAAP EPS of $21+.
Net sales of $4,313 million, up 61%, with full year 2026 net sales guidance raised to $14,650 million.
Q3 2026 revenue of $5.83 billion, up 83%, with 2027 revenue outlook of $40.0 billion.
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.