AI intelligence

Vercel introduced an integration for deploying long-running applications built with the OpenAI Agents API. OpenAI manages the agent loop and session state, while Vercel connects each session to an isolated Sandbox for execution and file access. The announcement describes persistent workspaces, signed webhooks and queues for reconnection, plus scaling to zero without a continuously running worker.

Vercel's Tako Search promotion ended September 30; the announcement says subsequent searches are billed at standard rates. The integration lets Gateway models query Tako's curated data and the live web, with domain and date filters. It supports citations and visualizations without a separate Tako account or key, and works across Gateway models without rebuilding the search integration.

Google announced football upgrades for Search on September 9, combining live game timelines, highlights and AI insights with richer player statistics. The initial professional-football feed launched on mobile in U.S. English. Fans can also connect Yahoo Fantasy or Sleeper accounts to AI Mode for roster-specific suggestions, including lineup and waiver decisions; that account-linking feature also launched in U.S. English.

A September 9 account from Google DeepMind explains how the short documentary Love, Rendered recreated an elderly couple's first meeting, which had never been recorded. Engineers combined restored photographs with performance capture based on the couple's present-day mannerisms, while Ethelle Shatz guided visual details. The result is a filmmaker-directed reconstruction of their recollections, with AI used to create younger likenesses and movement.

Mistral reports using AI agents to migrate 40,000 lines of a European energy operator's Fortran reservoir simulator into C++. The work combined numerical comparison tests, reconstructed documentation and human review of agent-generated changes. Mistral says fully autonomous attempts produced weaker results. The case study covers an initial portion of a 300,000-line system and cautions that projects without a working original program could be harder.

Vercel added persistent memory to eve agents, using named slots that define storage providers and sharing scopes, such as individual authenticated users. Relevant memories enter the model context before each turn; update behavior depends on the provider. On Vercel, the built-in file provider uses private Blob storage to preserve memories across restarts and deployments, with additional and custom providers supported.

AI Gateway custom costs now support cache-read and cache-write token rates. This lets custom cost metrics reflect negotiated cache pricing across providers. Add per_cache_read_token or per_cache_write_token to the cf-aig-custom-cost header: { "per_token_in": 0.000001, "per_token_out": 0.000002, "per_cache_read_token": 0.0000001, "per_cache_write_token": 0.0000005 } Cache-token pricing activates when either cache rate is present. An omitted cache rate defaults to per_token_in. If both cache rates are omitted, AI Gateway preserves the existing input and output calculation. Providers can include cache tokens within input tokens or report them separately. AI Gateway automatically accounts for these differences and prevents double-counting. For more information, refer to Custom costs.

Vercel added DeepSeek V4.1 Flash to AI Gateway with combined text and image input, including uses such as interpreting screenshots and charts. The announcement lists a one-million-token context window, responses up to 384,000 tokens, reasoning, tool use and prompt caching. Developers can select the model through Gateway or configure supported coding agents with the latest Vercel CLI.

Google DeepMind introduced AlphaGenome Atlas on September 8 with precomputed predictions for nine billion possible single-letter human DNA changes. A free portal makes the dataset available for academic research. Its AVI score combines AlphaGenome and AlphaMissense predictions to help researchers prioritize variants and inspect their predicted molecular effects. These are model-based research predictions; the announcement describes targeted experiments validating selected findings.

Mistral says it raised €3 billion in a Series D round at a valuation above €21 billion after the investment. Samsung Electronics led the round, with Scaleup Europe Fund and PSG Equity as co-leads. The company plans to expand AI research, computing infrastructure and international commercial operations, supporting its strategy of open-weight models and enterprise systems that customers can control.

Vercel added OpenAI's GPT Image 2.5 Flare and Sunburst to AI Gateway for image generation and editing from text prompts or reference images. Its announcement recommends Flare for faster generation and Sunburst when editing precision matters more. Vercel describes support for transparent backgrounds, complex layouts and targeted changes that preserve other image content, with both models available in its playground.

Sep 4, 22:26 UTC Resolved - On September 4, 2026, between 20:04 and 22:26 UTC, GitHub Copilot code review experienced an increased failure rate. Affected pull request reviews failed to complete or post review comments.

Vercel announced free access to inclusionAI's Ling 3.0 Flash Sante through October 4. The healthcare-focused model is presented for reasoning, research and evidence retrieval, with a 256K-token context and function calling. After the offer ends, the standard model endpoint continues at normal rates, while the endpoint ending in '-free' stops serving requests; free usage remains visible in traces.

Vercel's September 4 changelog added OpenAI's GPT 6 Astra to AI Gateway under the model identifier openai/gpt-6-astra. Vercel describes it as designed for extended agent tasks such as software navigation, data analysis and website development. The announcement provides an AI SDK example and coding-agent setup instructions, including links for Codex and Hermes, giving developers several ways to connect the model.

GitHub reports that a resolved September 3, 2026 incident disrupted Grok models in Copilot, including Grok 4.5 and 4.6, from 13:22 to 17:11 UTC. Requests to those models produced more errors, while other models remained unaffected. GitHub attributed the disruption to its upstream model provider, showed warnings inside affected products, and said service recovered after the provider applied a mitigation.

Google DeepMind and Google Research have introduced WeatherNext 3, which Google says uses live satellite observations to refresh global forecasts hourly. Temperature and moisture forecasts reach a five-kilometer grid, while other variables use coarser resolutions. Google says deployment begins across products including Search, Gemini and Maps, with data also available through cloud services. The announcement directs users to meteorological agencies for official warnings and public-safety advice.

Vercel announced on September 3 that Cursor Cloud Agents can use Vercel Sandbox as their execution environment. Cursor still operates the agent and inference loop, while each request gets an isolated Firecracker microVM for repository work and tests. The reference architecture adds session monitoring, retries and cleanup through Vercel services. Access to Cursor's Self-Hosted Machines APIs requires a Cursor Enterprise plan.

Google announced Fairwind, a restricted program giving selected government agencies, Google Cloud customers and cybersecurity partners access to Gemini 3.8 Flash Cyber paired with CodeMender. Google says the combination can identify vulnerabilities and generate validated fixes inside customers' cloud environments. Initial access prioritizes critical infrastructure and major software platforms, with participation subject to operational security requirements; the capability claims come from Google.

Google introduced Gemini 3.8 Flash for coding and agent workflows alongside Flash Cyber, a security-focused variant restricted to trusted defenders through Fairwind. Google reports improved vulnerability detection and patching performance for Cyber. The general Flash model is available through developer and enterprise products, but Google notes complex tasks can consume more tokens. The announcement's benchmark and capability results are vendor claims.

Google introduced Fairwind on September 2 as a restricted program for government agencies, selected Cloud customers and cybersecurity partners. It combines Gemini 3.8 Flash Cyber with CodeMender to identify vulnerabilities and produce validated fixes, according to Google. Initial access prioritizes public-sector systems, critical infrastructure and core technology platforms. Participants must restrict tool access to internal security teams and apply safeguards including multi-factor authentication.

Vercel added Google's Gemini 3.8 Flash to AI Gateway in its September 2 changelog. The listed configuration accepts text, images, PDFs and video, produces text, and supports tool use and web search. Vercel specifies a one-million-token context window and up to 65,536 output tokens, with thinking enabled by default. Developers select google/gemini-3.8-flash through the AI SDK or supported coding-agent integrations.

Google's September 1 roundup covers August launches including Gemini 3.7 Flash for coding and agents, Gemini 3.5 Transcribe for speech-to-text, and Pixel 11 devices. It also describes added task features in Gemini Live and Omni 1.1 Flash tools for extending clips, interpolating frames and 4K upscaling. These are Google's own product announcements, with some device features limited by country, language, age or subscription.

Google DeepMind introduced a video-analysis mode that lets Gemini search and re-examine relevant moments instead of processing footage at a fixed frame rate. The September 1 announcement covers Gemini 3.7 Flash, 3.6 Flash and 3.5 Flash-Lite through the Gemini API. Google reports lower token use and costs in its benchmarks, with particular benefits for long recordings; these are vendor-reported results.

Google announced Pics, a Nano Banana-based image creation and editing tool, with a rollout to AI Pro and Ultra subscribers and most Workspace business customers. Its September 1 post describes edits to individual objects, text changes and translation inside images, shared editing and multiple generated options. Docs and Slides integration began at launch, while Drive integration was planned for the following weeks.