The Deploy Log | Month by month
53 rows dated March 2026 are in The Deploy Log: 10 leads, 13 in Models and APIs, 21 in Tools and products, 1 in Robotics, hardware and chips and 8 in Breakthroughs, from 8 publishers. They include ChatGPT product discovery from OpenAI, Gemini 3.1 Flash Live from Google DeepMind, Voxtral TTS from Mistral, Forge from Mistral, GPT-5.4 mini and nano from OpenAI and 48 others. Every item sourced to the publisher's own page.
Visual product browsing, side-by-side comparison, and image-based search rolling out to all ChatGPT free, Go, Plus, and Pro users this week. Target, Sephora, Nordstrom, Lowe's, Best Buy, The Home Depot, and Wayfair are integrated.
Google's highest-quality audio model, with ComplexFuncBench Audio at 90.8% and Scale AI Audio MultiChallenge at 36.1% with thinking on. Gemini Live follows conversation for twice as long and Search Live expands to over 200 countries.
A 4B-parameter text-to-speech model with 70ms latency for a 10-second sample, 9 languages, and voice adaptation from a 3-second reference. Priced at $0.016 per 1k characters.
Forge introduced as a system for enterprises to build models grounded in proprietary knowledge. It supports pre-training, post-training, and reinforcement learning, plus dense and MoE architectures.
GPT-5.4 mini and nano released. Mini hits SWE-Bench Pro 54.4% and OSWorld-Verified 72.1%, running more than 2x faster than GPT-5 mini. Nano costs $0.20 per 1M input tokens and $1.25 per 1M output tokens.
ChatGPT now shows interactive visual explanations for more than 70 core math and science concepts, letting users adjust variables and see graphs change in real time. It is available globally across all plans.
next-forge 6 is a production-grade Turborepo template for Next.js apps. It makes Bun the default package manager, adds an installable agent skill, and makes every optional integration degrade silently when environment variables are missing.
ChatGPT for Excel is in beta, an Excel add-in that builds, updates, and analyzes spreadsheet models directly in workbooks. It is available globally to ChatGPT Business, Enterprise, Edu, Teachers, and K-12 users, and to ChatGPT Pro and Plus users outside the EU.
GPT-5.4 is released in ChatGPT as GPT-5.4 Thinking, the API, and Codex, with GPT-5.4 Pro also available. It scores 83.0% on GDPval, 57.7% on SWE-Bench Pro, 75.0% on OSWorld-Verified, and supports up to 1M tokens of context.
Gemini 3.1 Flash-Lite is rolling out in preview via the Gemini API in Google AI Studio and Vertex AI, priced at $0.25 per 1M input tokens and $1.50 per 1M output tokens. It achieves 86.9% on GPQA Diamond and 76.8% on MMMU Pro.
Mistral released Spaces, a CLI built for humans and agents, with every interactive input having a flag equivalent and context files for agent usability.
Lyria 3 Pro creates tracks up to 3 minutes long with intros, verses, choruses, and bridges, available in Vertex AI, Google AI Studio, the Gemini API, Google Vids, the Gemini app, and ProducerAI.
OpenAI published the backstory of its Model Spec, detailing the Chain of Command, authority levels, decision rubrics, and concrete examples that define intended model behavior.
ServiceNow AI released EVA, an end-to-end evaluation framework for voice agents that jointly scores accuracy (EVA-A) and experience (EVA-X), with an airline dataset of 50 scenarios and results for 20 systems.
Holotron-12B, post-trained from NVIDIA Nemotron-Nano-2 VL, raises WebVoyager from 35.1% to 80.5% and achieves over 2x higher throughput than Holo2-8B on a single H100.
Anthropic is committing an initial $100 million to the Claude Partner Network for training courses, dedicated technical support, and joint market development, with membership free of charge and applications open today.
Grok 4.20 is available on Vercel AI Gateway in Reasoning, Non-Reasoning, and Multi-Agent variants, with the Multi-Agent variant purpose-built for multi-agent orchestration and collaboration.
OpenAI is acquiring Promptfoo, an AI security platform trusted by over 25 percent of Fortune 500 companies, and will integrate its technology into OpenAI Frontier for automated security testing and red-teaming.
OpenAI describes social engineering-based prompt injection attacks and defenses like Safe Url that block silent data transmission.
OpenAI's Responses API now includes a shell tool and hosted container workspace for executing real-world tasks with filesystem, databases, and network access.
GPT-5.3 Instant is released as an update to ChatGPT's most-used model, reducing unnecessary refusals and improving web answer synthesis.
Modular Diffusers introduces a new way to build diffusion pipelines by composing reusable blocks, with a quickstart example running FLUX.2 Klein 4B and community pipelines including Krea Realtime Video achieving 11fps on a single B200 GPU.
Codex Security, an application security agent, is in research preview, having scanned 1.2 million commits in 30 days and found 792 critical findings.
The Custom Reporting API for AI Gateway is in beta for Pro and Enterprise plans, giving programmatic access to cost, token usage, and request volume across AI Gateway traffic including BYOK requests, with one platform saving over $80K by replacing a third-party proxy.
Notion 3.4 adds dashboard view for databases, a redesigned opt-in sidebar, presentation mode in beta for Plus, Business, and Enterprise, a tabs block, image generation, and page archiving.
The Vercel plugin now supports OpenAI Codex and Codex CLI, giving teams access to over 39 platform skills, three specialist agents, and real-time code validation.
MiniMax M2.7 is available on Vercel AI Gateway in standard and high-speed variants, with the high-speed variant delivering the same performance for 2x the cost at about 100 tokens per second.
Vercel Sandbox SDK 1.9.0 lets you set file permissions directly in writeFiles via a mode property, removing the extra chmod round-trip.
Streamdown 2.5 adds inline KaTeX support, staggered streaming animations with a default 40ms stagger, and fixes for code blocks, CSV exports, and Tailwind v3 compatibility.
Custom skills let you turn repetitive AI tasks into commands, usable from text selection or @mentions in agent chat. Skills are pages, so teams can share and build on them.
Chat SDK is a TypeScript library that builds bots for Slack, Teams, Google Chat, Discord, Telegram, GitHub, and Linear from one codebase, now with WhatsApp support.
Notion Academy videos and certifications are now live in Spanish, Portuguese, Japanese, Korean, French, and German.
AI Elements 1.9 adds a JSXPreview component that renders streaming JSX and closes unclosed tags, a PromptInput sub-component that captures page screenshots, and a Conversation download button.
Vercel Sandbox now supports creating Sandboxes with 1 vCPU and 2 GB of RAM for single-threaded or light workloads, with the default remaining 2 vCPUs and 4 GB.
Web Analytics and Speed Insights version 2 introduces resilient intake that dynamically discovers endpoints instead of relying on a single predictable path, available to all teams at no additional cost.
Workspace owners can now enforce a workspace-wide audio consent policy for AI Meeting Notes, playing a consent message at transcription start.
Notion AI can now search Salesforce accounts, leads, opportunities, and contacts alongside workspace content, via AI connectors in settings.
GPT-5.3 Chat (GPT 5.3 Instant) is now available on AI Gateway, focusing on tone, relevance, and conversational flow with fewer unnecessary refusals and caveats.
GPT-5.4 is now available in Notion's model picker, bringing faster responses and better token efficiency for demanding tasks.
GPT-5.4 and GPT-5.4 Pro are now available on AI Gateway, bringing agentic and reasoning leaps from GPT-5.3-Codex to all domains including knowledge work and coding.
Gemini 3.1 Flash Lite from Google is now available on AI Gateway, outperforming 2.5 Flash Lite on overall quality with notable improvements in translation, data extraction, and code completion.
Mercury 2 from Inception is now available on AI Gateway, delivering reasoning-grade quality at real-time latency for agentic loops, coding assistants, voice interfaces, and RAG pipelines.
Custom Agents now support MiniMax M2.5, an open weight model that is up to 10x more cost-efficient for basic tasks, available in the Custom Agent model picker.
Stripe is now generally available on the Vercel Marketplace and v0, letting you connect a production Stripe account and accept real payments.
LeRobot v0.5.0 adds full Unitree G1 humanoid support, Pi0-FAST autoregressive VLAs, Real-Time Chunking, streaming video encoding with zero wait between episodes, and EnvHub for loading simulation environments from the Hub.
Google Research's Vibe Coding XR uses Gemini with the XR Blocks framework to translate prompts into physics-aware WebXR apps in under 60 seconds, with a one-shot success rate around 70% initially, now improved after 11 releases.
Google Research introduced S2Vec, a self-supervised framework that learns embeddings of the built environment, performing best for zero-shot geographic adaptation in socioeconomic prediction, but weaker on environmental tasks.
Google Research introduced TurboQuant, a quantization algorithm that compresses key-value cache to 3 bits without accuracy loss, achieving up to 8x performance increase over 32-bit unquantized keys on H100 GPUs.
Groundsource uses Gemini to extract 2.6 million flood events from news reports across 150 countries, enabling urban flash flood forecasts.
Google's Flood Hub now provides urban flash flood forecasts up to 24 hours in advance, using a model trained on news-derived data.
A prospective study with Beth Israel Deaconess Medical Center found AMIE required zero safety stops and matched PCPs on diagnosis quality.
SpeciesNet is an open-source AI model that classifies 2,498 animal categories in camera trap images, trained on over 65 million labeled images, and it finds 99.4% of images containing animals with 94.5% of species-level predictions correct.
WAXAL is a large-scale, openly accessible speech dataset covering 27 Sub-Saharan African languages spoken by over 100 million speakers, with approximately 1,846 hours of transcribed natural speech for ASR and over 565 hours of high-fidelity recordings for TTS, released under CC-BY-4.0.
The edition every row came from arrives in your inbox, free.
Join free