NVIDIA released Nemotron-Labs-Diffusion, a single model that eliminates the need for a separate draft model in speculative decoding, achieving 6.82 accepted tokens per forward pass in self-speculation…
OpenAI launched ChatGPT Work on July 9, an autonomous agent that completes tasks like creating documents and web apps, while merging the standalone Codex app into a unified ChatGPT desktop application…
Gradium, a Paris-based voice AI startup backed by Nvidia, raised $100M to commercialize its audio-native language model that achieves sub-200ms latency, beating human conversation turn-taking baseline…
Chrome 149's WebMCP origin trial is now open for registration, and Microsoft Edge 147 has shipped native WebMCP support by default. WebMCP is a client-side protocol that allows websites to declare wha…
Anthropic restored access to Claude Fable 5, its most capable publicly available model, on July 1 after a 19-day suspension due to US export controls. The Mythos-class model leads benchmarks with a 1,…
MCP released beta SDKs for Python, TypeScript, Go, and C# on June 29, requiring migration before the July 28 deadline. The TypeScript SDK splits into separate client and server packages and drops Comm…
Lovable, a Swedish startup enabling non-coders to build web apps via natural language, is in talks to raise $300 million at a $13.2 billion valuation, doubling its valuation from seven months ago. The…
OpenAI launched GPT-5.6 on July 9 with three tiers—Sol, Terra, and Luna—each with distinct pricing and capability profiles, requiring developers to route tasks appropriately. Sol costs $5/$30 per mill…
Pydantic AI V2 went stable on June 23, 2026, introducing a capability-based architecture that bundles tools, lifecycle hooks, instructions, and model settings into composable units, replacing the flat…
Meta launched its paid AI API on July 9, offering the Muse Spark 1.1 model with a 1 million token context window and compatibility with OpenAI and Anthropic SDKs. The API costs $1.25 per million input…
Two developers, Michael Malis and Jason Seibel, used 17 AI coding agents to rewrite PostgreSQL in Rust, achieving 100% compatibility with all 46,066 regression tests of Postgres 18.3. The project, pgr…
Ollama raised $65M in Series B funding led by Theory Ventures, bringing total funding to $88M, and launched a cloud tier alongside its free open-source local AI tool. The company, used by 8.9 million …
Google released two generative media models via the Gemini API: Nano Banana 2 Lite, a fast image generator at $0.034 per image, and Gemini Omni Flash, a conversational video model at $0.10 per second,…
OpenAI released gpt-realtime-2.1 and gpt-realtime-2.1-mini voice models on July 6, featuring a 25% reduction in p95 latency through caching improvements and adding reasoning and tool use to the mini m…
Google released TabFM on July 1, a zero-shot foundation model for tabular data that generates predictions without training on the user's dataset, using in-context learning. The model achieves state-of…
Anthropic released Claude Opus 4.8 on May 28, featuring improved benchmarks and three quiet API changes that affect cost and capabilities for developers. The most significant update is support for mid…
Google released the Genkit Agents API on July 1, providing a stateful chat interface for building production-ready AI agents with built-in session management, interruptible tools for human approval, a…
Meta's custom Vistara CXL 2.0 memory expander ASIC, built in-house, is deployed across millions of production servers, repurposing DDR4 memory from decommissioned machines to expand capacity in DDR5-o…
OpenAI and Broadcom unveiled Jalapeño, OpenAI's first custom inference chip, on June 24, targeting 50% lower cost per token than Nvidia GPUs. The ASIC, built in nine months at TSMC's 3nm node, will no…
ZML launched LLMD, a free inference server that runs LLaMA, Gemma, Qwen, and Mistral models on NVIDIA, AMD, Google TPU, Intel, and Apple hardware from a single Docker image. Built in Zig and compiled …