GPT-6 Astra Ultrafast
GPT-6 Astra Ultrafast runs on NVIDIA Blackwell GPUs and is available now in the OpenAI API and to eligible ChatGPT Work and Codex users. It offers up to 8x faster token generation than the Astra Standard mode.
The Deploy Log | Week by week | Free, no signup
What shipped in AI, week of
308 rows are listed for in The Deploy Log, from 71 companies. Amazon Web Services shipped the most, 26 rows, then Anthropic with 23. Models first seen in the record this week: Claude Sonnet 5.5, Gemini 3.8 Flash Lite TTS, Gemini 3.8 Flash TTS, GPT-6 Astra UltraFast, GPT-6.1 Sol and 1 more. Every item sourced to the publisher's own page.
In figures
| Company | Rows |
|---|---|
| Amazon Web Services | 26 |
| Anthropic | 23 |
| Cloudflare | 18 |
| GitHub | 18 |
| Model Context Protocol | 17 |
| Google Cloud | 15 |
| OpenAI | 15 |
| Databricks | 13 |
| 12 | |
| Microsoft Azure | 12 |
| Figure | Value |
|---|---|
| Rows | 308 |
| Rows the week before | 134 |
| Change against the week before | +174 |
| Companies | 71 |
| Companies seen for the first time | 39: A2A Project, Agno, Amp, Arize, Baseten, Block, Cognition, CoreWeave, CrewAI, Deepgram, deepset, ElevenLabs, Firebase, GitLab, Hewlett Packard Enterprise, Kyutai, Lenovo, LibreChat, Liquid AI, LiveKit, Microsoft, Milvus, MLX-VLM, Model Context Protocol, Mozilla, Neon, Netlify, Perplexity, Prime Intellect, PyTorch, Qwen, Ray, Retool, Roboflow, Sentry, Shopify, Snowflake, Together AI and Ultralytics |
| Models seen for the first time | 6: Claude Sonnet 5.5, Gemini 3.8 Flash Lite TTS, Gemini 3.8 Flash TTS, GPT-6 Astra UltraFast, GPT-6.1 Sol and Qwen3.5 |
| Rows with no shipped date, placed by their edition or the day they were added | 18 |
| Lane | Rows | Rows the week before |
|---|---|---|
| Models and APIs | 31 | 9 |
| Tools and products | 268 | 121 |
| Robotics, hardware and chips | 3 | 1 |
| Breakthroughs | 2 | 0 |
| Funding and acquisitions | 2 | 0 |
Leads
Clef decision models
Cloudflare released two trained decision models, Clef and Clef-flash, hosted on Workers AI. Clef is currently the leader when evaluated against the Jev Decision Index, and the models are open-sourced on Hugging Face under an Apache 2.0 license.
Models and APIs
Qwen3.5-0.8B-Reverse-Text-RL
Prime Intellect released Qwen3.5-0.8B-Reverse-Text-RL, a short RL fine-tune for the reverse-text environment.
Qwen3.5-0.8B-Reverse-Text-SFT
Prime Intellect released Qwen3.5-0.8B-Reverse-Text-SFT, a short SFT fine-tune of Qwen3.5-0.8B.
MAI-Transcribe-2-Streaming
Microsoft launched MAI-Transcribe-2-Streaming, which debuts at no. 1 on Artificial Analysis, alongside MAI-Voice-2.1 and MAI-Voice-2.1-Flash for speech generation.
LFM2.5-350M-Diffusion-Exp
Liquid AI released LFM2.5-350M-Diffusion-Exp, a 350M block-diffusion model on Hugging Face with inference code and SGLang serving.
inclusionAI Ling 3.1 Flash
inclusionAI Ling 3.1 Flash is available on OpenRouter with 262K context and free pricing.
js-genai v2.27.0
js-genai v2.27.0 adds lyria-3.5, gemini-omni-1.1-flash, and gemini-omni-flash-preview to the Interactions Model enum.
python-genai v2.28.0
Adds lyria-3.5, gemini-omni-1.1-flash, and gemini-omni-flash-preview to the Interactions Model enum and supports continuation_token.
Deepgram Self-Hosted 261001
Self-hosted release 261001 ships Flux TTS watermarking, removes Whisper support, and adds FIPS 140-3 images.
Guided Vision in Gemini Live
Guided Vision launches in Gemini Live for Android.
Pareto 26.10 Preview
Pareto 26.10 Preview is available on OpenRouter at $0.80/$3.20 per 1M tokens with 1M context.
Show the other 21 rows
PixelUMM
NVIDIA released PixelUMM, a 15.2B-parameter encoder-free multimodal model for text, image, and video understanding and generation.
java-genai v1.75.0
v1.75.0 adds support for continuation_token in GenerateContent.
js-genai v2.26.0
js-genai v2.26.0 adds support for continuation_token in GenerateContent.
pplx-decider-v1-27b
Perplexity released pplx-decider-v1-27b, a decision model fine-tuned from Qwen3.8-27B, on Hugging Face.
python-genai v2.27.0
python-genai v2.27.0 adds support for continuation_token in GenerateContent.
Anthropic models in Seoul and Singapore
Anthropic models are available for in-region inference in Seoul and Singapore on Amazon Bedrock.
Claude in-country inferencing in India
Claude models are available for in-country inferencing in India on Amazon Bedrock.
Flux TTS Inline Controls
Flux TTS /v2/speak now supports inline pause (500-3000 ms) and pronunciation controls, with pause on batch and pronunciation on batch and streaming.
Gemini 3.8 Flash TTS models
Added Gemini 3.8 Flash TTS and Gemini 3.8 Flash Lite TTS models to the Python SDK.
Google Cloud Lakehouse public datasets
Adds Wikipedia pageviews and GitHub commit histories as public Apache Iceberg datasets in Google Cloud Lakehouse.
java-genai v1.74.0
Adds Gemini 3.8 Flash TTS and Gemini 3.8 Flash Lite TTS models, plus gcs_metrics_uri for tuning job metrics.
js-genai v2.25.0
Adds Gemini 3.8 Flash TTS and Gemini 3.8 Flash Lite TTS models to SDKs.
AdvancedMathBench-AutoVerifier
AutoVerifier model weights, tokenizer, and evaluation code released on Hugging Face.
GPT-6.1 Sol on Amazon Bedrock
GPT-6.1 Sol is now available on Amazon Bedrock.
NVIDIA Kumo Tabular
Open tabular foundation model now on Hugging Face in three sizes from 28M to 215M parameters.
OVIE-512
Kyutai released a 512x512 OVIE checkpoint with 41.6 ms per view on H100.
OVIE-ft-DL3DV
Kyutai released a DL3DV fine-tuned OVIE checkpoint with 0.1B params on Hugging Face.
Claude Sonnet 5.5 on OpenRouter
Claude Sonnet 5.5 is available on OpenRouter at $2/$10 per 1M tokens with 1.0M context.
Eleven v4 and Eleven v4 Turbo
Eleven v4 and Eleven v4 Turbo are available, with Turbo median inference latency of approximately 100 ms.
Grok 4.7 on Amazon Bedrock
Grok 4.7 is now available on Amazon Bedrock.
Mistral Python Client v3.0.0
Python client v3.0.0 released with 24 stable API surface removals.
Tools and products
GitHub Copilot CLI
Version 1.0.92-4 adds copilot config subcommands and improves MCP server startup responsiveness.
Ultralytics 8.4.173
Ultralytics 8.4.173 adds AMD Xilinx Vitis AI export for Versal AI Edge Gen 2 NPUs.
openai-java v4.76.0
openai-java v4.76.0 adds custom voices and agent session webhook events.
openai-node v7.28.0
openai-node v7.28.0 adds custom voice creation and agent session events.
supervision-0.30.7
supervision 0.30.7 fixes video processing hangs, dataset export, mAP, and Transformers loading.
AI Search
Cloudflare's AI Search is generally available, combining Workers AI, Vectorize, R2, and Browser Run into a fully managed index and retrieval pipeline, with billing starting .
Cloudflare Basin
Cloudflare Basin is generally available as a serverless data analytics platform built on Apache Iceberg and R2 Object Storage, including Basin Pipelines, Basin Catalog, and Basin SQL.
Workers KV Instant
Workers KV Instant, powered by Quicksilver, offers 100 times faster p99 reads than classic mode with reads resolving in under two milliseconds at p99, and writes replicating in around 250ms for 99% of writes.
Ollama v0.35.1
Ollama v0.35.1 supports Clef and Clef Flash, Cloudflare's new open-source decision models, through /v1/systemone, with Clef at 27B and Clef Flash at 9B, both multimodal.
Stateless installation tokens
The staged rollout of the stateless GitHub App installation token format is complete, with all newly minted tokens in the stateless ghs_APPID_JWT format at about 520 characters long instead of 40.
Show the other 258 rows
Laya on AI Gateway
Laya from Convai Innovations is free to use on AI Gateway through via Boundless, evaluating shared context against yes/no, choice, or scoring questions and returning structured answers with probabilities.
MAI models on AI Gateway
Vercel and Microsoft AI partnered to make MAI models available on AI Gateway, including MAI-Voice-2.1, MAI-Voice-2.1-Flash, and MAI-Transcribe-2 Streaming, billed at listed rates with no platform fee or markup on inference.
Protected Quick Tunnels
Starting with cloudflared 2026.9.3, you can add --allowed-mail to a Quick Tunnel command, and the tunnel only lets in the email addresses and domains you choose, with visitors proving ownership via a one-time PIN from Cloudflare Access.
Workers post-quantum crypto
Cloudflare Workers added support for post-quantum-resistant algorithms within Web Crypto, including ML-KEM-768 and ML-KEM-1024 for key encapsulation and ML-DSA-44, ML-DSA-65, and ML-DSA-87 for signatures, behind the webcrypto_modern_algorithms compatibility flag.
Claude Code v2.1.288
Claude Code v2.1.288 added $.ui.selection() for mods, a built-in gh api to cloud sessions, recovery for a prompt cleared with Ctrl+C, and --max-findings <n>|all to /code-review.
Claude Code v2.1.287
Claude Code v2.1.287 added Claude Mods, a built-in mod called You should know, an n:<text> filter to the agents view, and prompt_text to the OpenTelemetry user_prompt event.
vLLM v0.31.0
vLLM v0.31.0 adds a Transformers version upper bound in requirements.
Claude Agent SDK TypeScript v0.3.289
Updated to parity with Claude Code v2.1.289.
Claude Code v2.1.289
Fixes for deny rules, terminal freezing, symlink Read rules, and sign-out frequency.
Pydantic AI v2.54.0
Released pydantic-ai v2.54.0 with fixes for JSON schema, image generation, and durable execution.
@modelcontextprotocol/core@2.3.0
License field of package manifests is now Apache-2.0; LICENSE file carries full terms.
@modelcontextprotocol/express@2.0.2
requireBearerAuth and verifyBearerToken take a new optional expectedResource to accept only tokens issued for this resource.
@modelcontextprotocol/fastify@2.0.1
Server or McpServer now serves one connection at a time; stateless Streamable HTTP transport serves one request.
@modelcontextprotocol/hono@2.0.2
Server or McpServer now serves one connection at a time; stateless Streamable HTTP transport serves one request.
@modelcontextprotocol/server 2.3.0
requireBearerAuth and verifyBearerToken take a new optional expectedResource to accept only tokens issued for this resource.
@modelcontextprotocol/server-legacy 2.3.0
The license field of the package manifests is now Apache-2.0 with no code change.
AI Object Storage
CoreWeave launched AI Object Storage, an object storage service optimized for AI workloads.
Amazon EKS Kubernetes 1.37
Amazon EKS and EKS Distro now support Kubernetes version 1.37.
Android Device Streaming in Android CLI
Android Device Streaming is now available in Android CLI, giving agents access to real physical devices remotely.
Baseten CLI 1.1.0
Baseten CLI 1.1.0 adds model rename and volume sync, with sync supporting Hugging Face, S3, GCS, Azure, R2, or CoreWeave sources.
BigQuery Rust SDK
The Rust SDK for BigQuery is now generally available.
BigQuery Rust SDK GA
The Rust SDK for BigQuery is now generally available.
Cline Desktop v0.0.42
New /compact slash command and a redesigned Connectors tab with search and brand logos.
Cline SDK v0.0.90
Agent teams no longer slow down as a session runs or while a teammate is streaming; stream chunks and heartbeats are no longer persisted.
Cloudflare Web Search API
Web Search API beta on AI Gateway with Ceramic.ai, Exa, and Linkup providers, billed at list API price.
ComfyUI v0.38.2
ComfyUI v0.38.2 is released.
Data API builder 2.1.5
Adds SQL json and vector data type support to REST and GraphQL endpoints.
ElastiCache for Valkey OpenTelemetry
Amazon ElastiCache for Valkey now supports OpenTelemetry metrics and detailed monitoring.
Flutter replay masking
Session Replay for Flutter can mask selected text ranges like account numbers or email addresses while leaving safe labels visible.
GLM 5.2 on CoreWeave Inference
GLM 5.2 is now available on CoreWeave Inference.
GPT Live voice in Lovable
Lovable now lets users add live voice conversations to apps with GPT Live, working in TanStack Start apps and using credits.
GitHub Copilot CLI 1.0.92-3
Pre-release 1.0.92-3 adds a pre-conversation Ctrl+E environment picker and fixes input ordering, sandboxed shell network prompts, and Git credential handling.
GitLab AI Gateway Critical Patch
GitLab released AI Gateway versions 19.2.4, 19.3.2, and 19.4.1 with a critical security fix for self-hosted installations.
Google ADK Python v2.11.0
ADK Python v2.11.0 adds graceful cancellation, tool confirmation, a model consultation tool, and a built-in SQLite memory service.
Goose v1.53.0
Goose v1.53.0 adds support for Opus 5.5, GPT-6-sol, GPT-6-luna, and a lean ACP-only binary.
GuardDuty Runtime Monitoring
GuardDuty Runtime Monitoring is now included in the AWS Security Hub Threat Analytics plan.
JDBC Unity Catalog connections
JDBC Unity Catalog connections are now generally available on serverless compute, Databricks Runtime 19 and above, and SQL warehouses 2025.40 and above.
Langfuse v4.50.0
Langfuse v4.50.0 adds facet type, built-in protection, and rule idle time to evals, plus error IDs in tRPC toasts.
MCP TypeScript SDK 1.32.0
HTTP client transports now follow redirects only on the same origin unless redirectPolicy is set to follow.
MCP TypeScript SDK 2.3.0
Server.connect() now rejects while the instance is already connected, and a stateless server is one per request.
NVIDIA GB200 NVL72 Instances
CoreWeave is the first cloud provider to announce general availability of NVIDIA GB200 NVL72 instances.
NVIDIA HGX B200 Instances
CoreWeave expanded its NVIDIA Blackwell fleet with generally available NVIDIA HGX B200 instances.
NVIDIA RTX PRO 6000 Blackwell Server Instances
CoreWeave launched the first generally available NVIDIA RTX PRO 6000 Blackwell Server Instances.
Neon AI Gateway embedding models
Neon AI Gateway supports qwen3-embedding-0-6b and gte-large-en embedding models, returning 1024-dimension vectors.
ONNX Runtime v1.28.3
Patch release with memory-safety, model-validation, plugin execution, and packaging fixes.
OpenAI Agents Python v0.23.0
Adds configurable MCP listing page limits, opt-in Docker removal protection, memory consolidation turns, and encrypted history scan budget.
OpenAI models in Retool
Retool supports GPT 6 Luna, GPT 6 Sol, and enterprise-only GPT Image 2.5 Flare and Sunburst for AI resource queries.
OpenFeature provider for Elixir
Elixir applications can evaluate Feature Flags through a new optional OpenFeature provider.
OpenJev (Qwen3.5 4B)
OpenJev (Qwen3.5 4B) is available as a pay-per-token model service on Unity Gateway.
Pydantic AI v2.53.0
Fixes a high-severity concurrency slot leak in ConcurrencyLimitedModel.
Ray 2.59.0
Ray Data LLM and Ray Serve LLM APIs are GA, with vLLM 0.27.0 and an external disk-backed shuffle backend.
React Native fatal error recovery
Error Tracking in React Native preserves fatal JavaScript exceptions through process crashes for reporting after relaunch.
Retool natural language resources
Prompt the app building agent to create REST API or PostgreSQL resources directly from the app builder.
Roboflow Inference v1.7.3
RF-DETR Nano on Mac goes from 72 ms to 16 ms end to end.
Ruby MCP analytics
Experimental MCP analytics is available for Ruby servers built with the Ruby MCP gem.
SecurityAdvisory GraphQL API fields
The SecurityAdvisory object gained five new fields and the securityAdvisories query gained two new filters.
Spanner queues
Spanner queues are generally available, providing native transactional messaging embedded directly within Spanner.
Strict service token authentication
Cloudflare Access added a strict service token authentication setting, default on for organizations created after .
SwiftUI tap autocapture
The iOS SDK can now autocapture SwiftUI control taps with an opt-in configuration.
Workers KV namespace jurisdictions
Workers KV namespace jurisdictions are generally available with eu, us, and fedramp regions.
Workers Observability in Custom Dashboards
Cloudflare adds Workers Observability Logs and Traces datasets to Custom Dashboards, with up to 100 dashboards per account.
openai-python v3.24.0
openai-python v3.24.0 adds custom voice creation and agent session events.
Agno v3.1.0
Agno v3.1.0 adds role-based access control and a filesystem to AgentOS.
Azure Cosmos DB Shell
Azure Cosmos DB Shell is now available in the Azure portal as a preview, accessible from Data Explorer.
Azure Pipelines GitHub-hosted agents
GitHub-hosted agents with pay-as-you-go pricing are generally available, including macOS SKUs and Linux/Windows previews.
BigQuery conversational analytics
Chatting with graphs in conversational analytics is now generally available, using multiple graphs, tables, views, and UDFs as data sources.
BigQuery metadata enrichment
BigQuery pipelines now supports automated metadata enrichment and Knowledge Catalog data quality scorecard integration, generally available.
Claude Agent SDK TypeScript v0.3.287
Adds optional remote-session latency fields and fixes partial message streams and revoked login errors.
Claude Opus 5.5 in Retool
Claude Opus 5.5 is available in Retool through Anthropic and Amazon Bedrock for Enterprise AI resource queries.
Copilot CLI 1.0.92-0
Copilot CLI 1.0.92-0 fixes MCP tools continuing to work after OAuth reauthentication when tool definitions are unchanged.
CoreWeave AI Object Storage
Cross-region writes and Archive storage are now available in CoreWeave AI Object Storage.
Databricks IP Functions
IP Functions for network analytics are generally available on the Lakehouse.
DigitalOcean Inference and Agents balance
Prepaid balance and auto-reload for Serverless Inference and Managed Agents, with monthly plans in public preview.
Dynamic workflows in Copilot
Dynamic workflows are now available in Copilot CLI, the GitHub Copilot app, and the GitHub Copilot SDK.
GitHub Actions macOS 14 retirement
The macOS 14 runner image will be retired on , with brownout periods starting .
GitHub Copilot CLI 1.0.91
Adds sandbox CA commands for check, create, trust, rotate, and remove, plus session timeline and Windows sandbox fixes.
GitHub Copilot CLI 1.0.91-1
Adds copilot sandbox ca commands for proxy CA trust and fixes shutdown telemetry flush.
GitHub Copilot computer use
Computer use is now available in public preview in GitHub Copilot CLI and the GitHub Copilot app on macOS and Windows.
GoldenGate Veridata connections
GoldenGate Veridata adds support for 15 new connection types including Amazon Aurora, Azure Database, and SingleStoreDB.
Haystack 3.3.0
Haystack 3.3.0 released with faster SentenceWindowRetriever, BM25 retrieval changes, and CJK tokenization fixes.
Haystack v3.3.0
SentenceWindowRetriever now queries the Document Store once per run instead of once per retrieved document.
Hugging Face Hub v2.1.0
Hugging Face Hub v2.1.0 adds Jobs retries, rescheduling, port exposure, ZeroGPU quota tracking, and a revamped Model Catalog.
Kilo Code v7.8.3
Kilo Code v7.8.3 adds a copy content button, document context menu actions, a manual Fetch Models button, and multi-root workspace file suggestions.
LTAP Direct Writes
LTAP Direct Writes is generally available for Lakebase, writing initial loads and full refreshes directly to storage.
LibreChat v0.8.8
LibreChat v0.8.8 adds agent control, unified agent builder, and background tasks.
Meet-Teams Android Interop
Google Meet and Microsoft Teams interoperability on Android (AOSP) devices is now generally available.
NVIDIA DOCA Agent Skills
DOCA AI agent skills for Flow, GPUNetIO, PCC, and RDMA are available on GitHub.
OCI DRG NAT
OCI Dynamic Routing Gateway now supports NAT for IPv4 addresses, configurable on DRG attachments.
OpenAI Go SDK v3.69.0
Adds typed output schema binding, translation and session trace APIs, and requires secure credential transport.
OpenAI Node SDK v7.26.0
OpenAI Node SDK v7.26.0 adds agent session trace listing and typed application functions for Agents tools.
Polaris 2.0.0-rc.1
Polaris 2.0.0-rc.1 updates components to the new admin style and improves mounting and update performance.
PostHog HogQL batch export modifiers
HogQL-model batch exports can now use their own HogQL modifiers when created or updated.
Redshift Cross-Region Queries
Amazon Redshift now supports cross-Region queries for your data lake.
Shopify Next Gen Events
Next Gen Events is generally available across 18 topics with the 2026-10 API version.
SubscriptionContractCalculation API
Shopify released a SubscriptionContractCalculation API that runs contract edits through checkout, supporting bundles and Shopify Functions.
Ultralytics v8.4.171
Adds end-to-end AMD GPU support via ROCm and MIGraphX for training and native PyTorch mode.
Workers Web Crypto ML-KEM/ML-DSA
Workers Web Crypto API now supports ML-KEM-768, ML-KEM-1024, ML-DSA-44, ML-DSA-65, and ML-DSA-87.
Workspace Client-side encryption simple setup
A new simple setup path in the Admin console lets admins set up Client-side encryption in minutes using Cloud HSM Keys and Google Identity.
livekit-agents@1.8.4
livekit-agents 1.8.4 adds smallestai TTS continuations, baseten inline instructions, and Azure Voice Live Realtime API support.
openai-python v3.23.0
openai-python v3.23.0 adds typed answers from session streams, session traces, Realtime translations, and typed application actions as agent tools.
openai-ruby v0.98.0
openai-ruby v0.98.0 adds typed output parsing, hosted files, session traces, and beta agent tool binding.
5th Gen AMD EPYC on Aurora and RDS
Amazon Aurora and RDS now support 5th generation AMD EPYC processors.
AI Gateway Browserbase Tools
Browserbase Search and Fetch tools are available through AI Gateway, requiring AI SDK 7.0.116 or later.
AI Gateway Machine Payments
AI Gateway supports Machine Payments in beta for /ai/run with select open models via x402 stablecoin payments.
AWS CLI Agent Toolkit updates
AWS CLI now supports bulk skill updates and version checks for the Agent Toolkit for AWS.
AWS Continuum for Penetration Testing
AWS Continuum for Penetration Testing is now available in 6 additional regions.
AWS Jam AI Agent Experiences
Three new AWS Jam experiences for building, deploying, and hardening AI agents are now available.
Amazon Managed Grafana 13.2
Amazon Managed Grafana now supports creating Grafana 13.2 workspaces.
Amazon S3 Vectors metadata pre-filtering
Amazon S3 Vectors now supports metadata pre-filtering for higher recall on filtered searches.
Anthropic Java SDK v2.67.0
Adds refusal stop reason, Claude Enterprise analytics, spend limits, RBAC, per-user usage reports, and Plugins to the Admin API.
Anthropic Python SDK v1.10.0
Anthropic Python SDK v1.10.0 adds Admin API analytics, spend limits, RBAC, and Plugins.
Aurora Serverless faster scaling
Amazon Aurora serverless now scales faster to support agentic AI and other bursty workloads.
Aurora and RDS R8a instances
Amazon Aurora and RDS now support AMD-based R8a instances.
Bedrock Claude in-region London
Amazon Bedrock expands Claude models in-region support in the UK (London).
Claude Code v2.1.286
Claude Code v2.1.286 adds a count to stacked permission prompts and mouse support for "N more" rows in fullscreen lists.
Cline CLI v3.0.66
CLI binaries now built with Bun 1.4.2 and Windows proxy startup fixed.
Cline Desktop v0.0.38
Linux ARM64 .deb and .rpm packages now ship alongside x64 builds.
Cline Desktop v0.0.39
Desktop v0.0.39 fixes blank window on Linux with NVIDIA driver and adds app info to .deb and .rpm packages.
Cline SDK v0.0.88
Anthropic requests now use server-side refusal fallback and provider allow_fallbacks; Stop during turn preparation no longer dropped.
Cline v4.1.22
Adds Bee and Pareto Inference providers; GPT-6.1 Sol becomes default model for OpenAI and other providers.
Cloudflare One Client for Windows
Beta release 2026.8.2033.1 is available with fixes for traffic blocking, reauthentication, DNS, and client crashes.
Copilot Code Reviews for Azure Repos
Effort levels, comment resolution, and audit log events added to Copilot Code Reviews for Azure Repos.
Cosmos DB Mirroring VNet Data Gateways
Cosmos DB mirroring in Fabric now supports virtual network data gateways for accounts secured by private endpoints or virtual networks.
Data Agent Kit GA
Data Agent Kit is generally available with support for BigQuery Graph, Bigtable, and Managed Service for Apache Spark.
Docs, Sheets, Slides comments API
Google Docs, Sheets, and Slides APIs now support programmatic comments; Docs adds suggested edits.
GPT-6 Astra UltraFast on Bedrock
OpenAI GPT-6 Astra now supports UltraFast mode on Amazon Bedrock.
GPT-6.1 Sol on DigitalOcean Inference
GPT-6.1 Sol is now available on DigitalOcean Inference for serverless inference and Agent Development Kit.
GitHub Copilot CLI 1.0.91-0
Version 1.0.91-0 lets statically analyzable read-only shell pipelines enter execution-evidence review; incomplete or unbound pipelines require explicit approval.
GitHub Copilot CLI v1.0.90
Release 1.0.90 adds GPT-6.1 Sol model selection, --mcp-github-auth, and session-scoped read-only directory approvals.
GitHub Copilot CLI v1.0.90-6
Added GPT-6.1 Sol to model selection.
Google Cloud CLI remote MCP server
Google Cloud introduced the Google Cloud CLI remote MCP server in preview, giving AI agents access to gcloud and bq commands.
Google Vids caption customization
Google Vids captions can now be customized for font, position, and animation.
HSTU Generative Recommender Workflow
NVIDIA recsys-examples now provides an end-to-end HSTU inference workflow with Dynamo-Triton, achieving up to 5.93x lower latency at batch size 8.
Lakebase Postgres extensions
Supported Postgres extensions updated to newer versions, including postgis 3.6.4 on Postgres 18 and hll 2.21.
Ling 3.1 Flash on AI Gateway
Ling 3.1 Flash from InclusionAI is available on AI Gateway, free through , with 262K-token context.
MCP Inspector 2.2.0
MCP Inspector 2.2.0 released with sandbox proxy loading and keychain degradation fixes.
MCP Inspector 2.3.0
MCP Inspector 2.3.0 released with loopback binding and sandbox port pinning fixes.
MCP Inspector 2.4.0
MCP Inspector 2.4.0 released with app-rendered form elicitations and SecretStore support.
MCP Inspector 2.5.0
MCP Inspector 2.5.0 released with Firefox smoke tests and tool schema rendering fixes.
Microsoft plugin registry
Plugin registry rolling out now for Copilot Home, Code, Autopilot, Microsoft 365 apps, and SharePoint.
NVIDIA cuObject and SCADA Server SDK
NVIDIA cuObject client and server libraries are now generally available, with the SCADA Server SDK enabling storage providers to build servers.
Nemotron 3 Diarization
Transformers v5.18.0 adds Nemotron 3 Diarization, an open-weight streaming speaker diarization model supporting up to eight speakers.
RDS M8a instances
Amazon RDS now supports AMD-based M8a instances.
Raycast v2.6
Raycast v2.6 reduces host memory by 50% and adds Auto model preferences: Lightweight, Balanced, Flagship.
Retool 4.66 Edge
Retool 4.66 is now available on the Edge release channel for self-hosted organizations.
SDK v0.0.89
Oversized MCP and Composio tool results can now be recovered in full via a per-session cache with a cline://cache URI.
Spanner Omni GA
Spanner Omni, the deploy-anywhere version of Spanner, is now generally available for on-premises and other clouds.
Together AI LoRA rank limit
Together AI now allows LoRA adapters with rank up to 128, up from 64.
Together Link beta
Together Link beta runs six coding agents on Together AI models on macOS and Linux, installable with one command.
TorchVision 0.29.1
TorchVision 0.29.1, a patch release compatible with PyTorch 2.14.1, with no new features.
Vercel Sandbox Secure Compute
Vercel Sandbox now supports Secure Compute, connecting sandboxes to a team's dedicated network with static IPs and VPC peering.
WorkSpaces Core Blackwell GPU
Amazon WorkSpaces Core Managed Instances adds support for NVIDIA Blackwell GPU.
anthropic-sdk v0.130.0
Adds refusal stop reason, Claude Enterprise analytics, spend limits, RBAC, usage reports, and plugin marketplaces to the Admin API.
anthropic-sdk-go v1.78.0
anthropic-sdk-go v1.78.0 adds a list spend limits endpoint and marks Claude Sonnet 4.5 deprecated with end-of-life November 30, 2026.
anthropic-sdk-java v2.68.0
anthropic-sdk-java v2.68.0 adds a list spend limits endpoint and marks Claude Sonnet 4.5 deprecated with end-of-life November 30, 2026.
langchain-openai 1.6.7
langchain-openai 1.6.7 released with Azure workload identity discovery.
openai-java v4.73.0
Adds external storage provider callback handlers, session traces, translation secrets, and pagination and Azure deployment fixes.
sdk v0.131.0
Adds list spend limits endpoint and deprecates sonnet 4.5.
AI Gateway User Insights
User Insights for AI Gateway is available to all customers at no additional cost, showing model overuse and potential savings.
Agent Runners link previews
Every site created with Agent Runners now includes Open Graph tags for rich link previews.
Azure Cosmos DB in Fabric Database Hub
Azure Cosmos DB accounts can now be discovered and monitored from Fabric Database Hub in public preview.
BigQuery FAILOVER_HISTORY view
BigQuery users can query INFORMATION_SCHEMA.FAILOVER_HISTORY for near real-time failover events.
Claude Code v2.1.285
Claude Code v2.1.285 adds CLAUDE_CODE_DISABLE_WEB_FETCH, claude --desktop, and plugin configure options.
Claude on Bedrock in India, South Korea, Singapore
Amazon Bedrock expands Claude model availability to India, South Korea, and Singapore.
Cloudflare Application Profiles
Positive security enforcement with Cloudflare Application Profiles.
Cloudflare Containers disk limit
Custom Container instance types can now allocate up to 20 GB disk regardless of memory.
Cloudflare network log fields
Gateway network logs and Network Session Logs now identify Mesh, Workers VPC, and Tunnel replica traffic.
ComfyUI v0.38.0
ComfyUI v0.38.0 released.
DOKS Isolated Worker Nodes
Isolated Worker Nodes for DigitalOcean Kubernetes are generally available, removing public IPv4 addresses from nodes.
Databricks JDBC Driver (Simba) 2.8.4
Simba JDBC Driver (Legacy) 2.8.4 adds Java 25 support and upgrades libthrift, httpclient5, httpcore5, and lz4-java.
DiskANN Vector Index in Azure SQL
DiskANN Vector Index and Search are generally available in Azure SQL Database, Managed Instance, and Fabric SQL database.
ElastiCache Serverless for Valkey
Amazon ElastiCache Serverless for Valkey now supports public endpoints.
Final Cut Camera 2.4
Final Cut Camera 2.4 adds variable aperture on iPhone 18 Pro, Cinematic mode effects for regular video, and iOS 27 pro options.
Firefox FX 157
Firefox FX 157 rolls out the new design to all desktop and mobile users.
GKE Agent Sandbox for RL
GKE Agent Sandbox optimized for RL and Agent Sandbox RL orchestration SDK are generally available, with 10x-45x faster time-to-first-command.
GPT-6.1 Sol in GitHub Copilot
GPT-6.1 Sol is generally available in GitHub Copilot for Pro+, Max, Business, and Enterprise users.
GPT-6.1 Sol on AI Gateway
GPT-6.1 Sol is available on Vercel AI Gateway as openai/gpt-6.1-sol with lower standard input and output pricing than GPT-6 Astra.
GPT-6.1 Sol on Netlify AI Gateway
OpenAI's GPT-6.1 Sol model is available through Netlify's AI Gateway and Agent Runners with zero configuration required.
Gemini 3.8 Flash Lite TTS in Vids
Google Vids upgraded AI voiceovers and avatar narration to Gemini 3.8 Flash Lite TTS, available now.
Gemini 3.8 Flash default for AlphaEvolve
Gemini 3.8 Flash is now the default model for AlphaEvolve experiments, replacing Gemini 3.5 Flash.
Gemini Enterprise security agents
New partner-built security agents and AI defenses from Acalvio and Britive are available in Gemini Enterprise.
GitHub Copilot CLI v1.0.90-5
Fixed model availability message and MCP tool calls completing despite progress updates.
Jev in LangSmith Evals
Jev is available as a judge for evaluations in LangSmith, offering fast, low-cost evaluation of open-ended agent behavior.
Lasv5 and Laosv5 Azure VM series
Storage optimized Lasv5 and Laosv5 Azure VM series are generally available.
NVIDIA VSS Blueprint 3.3
VSS Blueprint 3.3 adds the Build Vision Agent skill and Adaptive EVS, cutting alert contextualization latency 17% and increasing concurrent VLM streams 46%.
Ollama v0.35.0
Ollama now supports decision models through /v1/systemone, returning choices, probabilities, and scores.
OpenAI GPT-6.1 Sol on Unity Gateway
OpenAI GPT-6.1 Sol is available on Unity Gateway via Foundation Model APIs.
OpenAI Python SDK v3.21.0 adds GPT-6.1 Sol model identifier
The openai-python SDK release adds the GPT-6.1 Sol model identifier.
Plaid Speed
Plaid inference for GPT-6 Astra modes runs up to 6x faster and costs 6x per token.
Qwen Code Desktop v0.24.7
Qwen Code Desktop v0.24.7 released with session diagnostics fix, Java SDK attestation client, trajectory time range, and Qwen Live endpoint settings.
RunPod Deploy When Available
RunPod's Deploy When Available is now GA, queuing for any GPU spec and deploying when capacity opens.
Seer Agent
Seer Agent gains new capabilities for interacting with Sentry projects and collecting data from more places.
Sentry JavaScript SDK 11.1.0
Sentry JavaScript SDK 11.1.0 traces TypeSafe Jev and adds typesafeIntegration for @typesafe-ai/sdk, enabled by default.
Shopify App Pricing migration
Apps using the Billing API can now migrate existing subscriptions to Shopify App Pricing via Partner Dashboard tool or Shopify CLI commands.
TESSERA Earth observation embeddings
TESSERA global Earth observation embeddings are now available on AWS Open Data.
Threat Signals
Agentic skills for open-source threat intelligence, free for every Cloudflare account.
TogetherLink beta
TogetherLink beta runs Claude Code, Codex, and desktop apps on Together AI models with automatic routing, cost tracking, and headless mode.
Visual Studio BYOM
Bring Your Own Model is available in preview and enabled by default across Community, Professional, and Enterprise SKUs.
WSL Containers (WSLC)
WSL containers is now generally available with the new wslc.exe command.
a2a-python v1.2.0
a2a-python v1.2.0 adds HTTP+JSON responses, caching headers, multi-replica cluster mode, and media type validation.
langchain-anthropic 1.7.5
Release adds Claude Sonnet 5.5 compatibility and serializes invalid tool calls as tool use on replay.
langchain-core 1.6.6
langchain-core 1.6.6 released with Claude Sonnet 5.5 compatibility fix.
openai-go v3.68.0
OpenAI Go SDK v3.68.0 adds computer use to beta agents.
openai-java v4.72.0
GPT-6.1 Sol model identifier and computer use for beta agents in Java SDK v4.72.0.
openai-node v7.25.0
OpenAI Node SDK v7.25.0 adds computer use to beta agents.
1Password Event Logs connector (Beta)
1Password Event Logs connector is now available in Lakeflow Connect in Beta, ingesting 1Password event logs into Databricks.
@modelcontextprotocol/core 2.2.0
@modelcontextprotocol/core 2.2.0 published with deprecation warning for OAuth providers missing expectedIssuer.
@modelcontextprotocol/server 2.2.0
@modelcontextprotocol/server 2.2.0 published with fixes for unhandled promise rejection and stack overflow.
@modelcontextprotocol/server-legacy 2.2.0
@modelcontextprotocol/server-legacy 2.2.0 published with dependency update to core 2.2.0.
Agent Runners inline preview
Agent Runners now shows a deploy preview inline next to the prompt.
Akamai WAF connector (Beta)
Akamai WAF connector is now available in Lakeflow Connect in Beta, ingesting WAF security events from Akamai into Databricks.
Anthropic Python SDK v1.9.0
Adds claude-sonnet-5-5, between_tools thinking type, and cache diagnostics GA.
Ask in Google Earth Engine
Ask, a Gemini-powered chat assistant, is available in Google Earth Engine Code Editor for writing, debugging, and optimizing geospatial queries.
BigQuery Data Transfer Service MCP server
The BigQuery Data Transfer Service MCP server is now Generally Available (GA).
Blazor AI components
Experimental Blazor AI components for agentic UI, available as a prerelease package with .NET 11 RC1 SDK.
Claude Agent SDK Python v0.2.161
Python SDK v0.2.161 released with bundled Claude CLI updated to version 2.1.284.
Claude Code v2.1.284
Claude Sonnet 5.5 is now the default Sonnet model on the Anthropic API with 1M context, $2/$10 per Mtok, $0.20/Mtok cache reads.
Claude Sonnet 5.5 in Devin
Claude Sonnet 5.5 is available in Devin, scoring 64.4% on FrontierCode 1.1.
Claude Sonnet 5.5 in GitHub Copilot
Claude Sonnet 5.5 is generally available in GitHub Copilot for Pro, Pro+, Max, Business, and Enterprise users.
Claude Sonnet 5.5 on AI Gateway
Claude Sonnet 5.5 is available on AI Gateway as anthropic/claude-sonnet-5.5.
Claude Sonnet 5.5 on AWS
Claude Sonnet 5.5 is now available on AWS.
Claude Sonnet 5.5 on AWS GovCloud (US)
Claude Sonnet 5.5 is now available on AWS GovCloud (US).
Claude Sonnet 5.5 on DigitalOcean Inference
Claude Sonnet 5.5 is now available on DigitalOcean Inference for serverless inference and Agent Development Kit.
Claude Sonnet 5.5 on Netlify
Claude Sonnet 5.5 is available through Netlify AI Gateway and Agent Runners with zero configuration required.
Codex CLI rust-v0.158.0
Codex CLI 0.158.0 adds copy-on-select, OAuth client secrets for MCP, bearer tokens for exec-server, and transparent image backgrounds.
CrewAI 1.15.23
CrewAI 1.15.23 adds native Gemini 3.8 Flash support and crewai eval improvements.
CrowdStrike Falcon Event Stream connector
CrowdStrike Falcon Event Stream connector is available in Lakeflow Connect in Beta, ingesting Falcon Event Stream events into Databricks.
Firebase Phone Number Verification
10 new carrier networks are available today across Malaysia, Spain, Pakistan, and six Vodafone countries.
Firebase plugin for Codex
Firebase agent plugin is available in the Codex plugin directory with agent skills, MCP server, and CLI tools.
GitHub Copilot CLI 1.0.89
Adds Claude Code rule files as custom instructions, click-to-focus inputs, unread turn dots, routing tier suggestions, and MCP OAuth scope support.
GitHub Copilot CLI v1.0.89-6
Copilot CLI v1.0.89-6 improves PR creation to follow repository pull request templates and fixes shell output metadata.
Jev-as-a-Judge in Arize AX
Arize AX now integrates directly with TypeSafe AI, bringing native Jev-as-a-Judge evaluations into the Arize evaluation workflow.
Luxembourg Azure Extended Zones
Luxembourg Azure Extended Zones are generally available.
MCP TypeScript SDK 1.31.0
Stored OAuth tokens and client information now include an issuer field; pass expectedIssuer when constructing OAuth providers.
MCP TypeScript SDK 2.2.0
MCP TypeScript SDK 2.2.0 released with expectedIssuer required for machine-to-machine OAuth providers.
MLX-VLM v0.7.4
Generate multiple images per prompt via num_images for qwen_image, flux2, ideogram4, plus audio and padding fixes.
Microsoft 365 Unified Audit Logs connector
Beta connector ingests unified audit events from Microsoft 365 workloads into Databricks.
Milvus v2.6.24
Milvus v2.6.24 released with lower peak memory and faster filtered search.
Milvus v3.0.2
Milvus v3.0.2 released with search and query performance improvements.
Okta System Logs connector
Beta connector ingests Okta System Log audit and security events into Databricks.
OpenAI Python SDK v3.20.0
OpenAI Python SDK v3.20.0 adds Agents credential and session options, Cyber access programs, and incremental WebSocket snapshots.
PostHog Backfill App Store analytics
Data Warehouse can backfill App Store Connect analytics from account inception or January 2024 on first sync, resync, or full refresh.
Rekognition Face Liveness Feedback Codes
Amazon Rekognition Face Liveness now returns Feedback Codes.
SQLAlchemy 2.1 mssql-python support
SQLAlchemy 2.1.0 is generally available with built-in mssql-python support for SQL Server and Azure SQL.
Valkey 9 on DigitalOcean
Valkey 9 is now available for database clusters, and new clusters use it by default.
Vector search in Azure SQL
Vector search and vector indexes are generally available in Azure SQL.
Workday Activity Logging connector
Beta connector ingests Workday user activity log events into Databricks.
Z4D machine family GA
Storage-optimized Z4D VM and bare metal instances are generally available with up to 84,000 GiB Local SSD and up to 40% better performance than Z3.
anthropic-sdk-go v1.76.0
Anthropic Go SDK v1.76.0 adds claude-sonnet-5-5, between_tools thinking type, and cache diagnostics GA.
anthropic-sdk-java v2.66.0
Anthropic Java SDK v2.66.0 adds claude-sonnet-5-5, between_tools thinking type, and cache diagnostics GA.
bedrock-sdk v0.34.0
Anthropic Bedrock SDK v0.34.0 adds claude-sonnet-5-5.
claude-agent-sdk-typescript v0.3.284
claude-agent-sdk-typescript v0.3.284 adds renamed skill directory aliases, ultracode settings, and fixes forkSession and getSessionMessages bugs.
foundry-sdk v0.5.0
Foundry SDK v0.5.0 adds claude-sonnet-5-5.
langchain 1.4.3
langchain 1.4.3 supports Bedrock Mantle chat models in init_chat_model and recognizes GPT-6 structured output.
langchain-fireworks 1.7.0
langchain-fireworks 1.7.0 adds prompt caching middleware and fixes mid-stream read timeouts.
playwright-mcp v0.0.83
Bug fixes for browser_find, browser_navigate, browser_reload, and download handling.
sdk v0.129.0
SDK v0.129.0 adds claude-sonnet-5-5, between_tools thinking type, and ClientToolUnion.
vertex-sdk v0.20.0
Vertex SDK v0.20.0 adds claude-sonnet-5-5.
Robotics, hardware and chips
DGX Spark 64GB
NVIDIA DGX Spark will be available with 64GB of unified memory from Acer, ASUS, Dell, Gigabyte, HP and MSI starting Friday, at $4,999, supporting up to 100-billion-parameter models fully on device.
Lenovo AI Express and AI Fast Start
Lenovo AI Express quick-start program and AI Fast Start pilot service, with order-to-ship in as little as 15 business days and working pilot in 90 days.
Vera Rubin NVL72 on CoreWeave
CoreWeave announced availability of NVIDIA Vera Rubin NVL72 systems with Spectrum-X 102.4T Ethernet networking, and Cognition is running production workloads.
Breakthroughs
AstaBrief 8B
Ai2 open-sourced AstaBrief 8B, a model that turns a research question and retrieved literature excerpts into a cited report, with Fast mode averaging 51.1 seconds per report compared with 178.5 seconds for Thinking mode, about 3.5x faster.
Olmo-core 3
Ai2 released Olmo-core 3, an open MoE training system benchmarked at over one trillion total parameters, with a 47-billion-parameter MoE processing 52,000 tokens per second per GPU on eight NVIDIA B300 GPUs, about 2.7x the throughput of the earlier implementation.
Funding and acquisitions
Snowflake $3.5B Convertible Notes
Snowflake proposes a private placement of $3.5 billion of 0.00% convertible senior notes.
HPE Networking AI-era guidance
HPE expects Data Center Networking revenue to grow at a low-to-high 50s percent CAGR through fiscal 2029.
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.