Perplexity Engineering detailed its GPU embedding serving stack for pplx-embed, revealing that it reuses prefill and decode kernels from its LLM engine rather than building a separate embedding engine…
GitHub has released Project HydraFusion, a research preview in GitHub Copilot CLI that builds a per-request execution plan across multiple models and providers, selecting from Single, Cascade, or Crit…
Nous Research has added a one-click local model setup flow to Hermes Desktop, its MIT-licensed, free desktop build of the open-source Hermes Agent, which automatically reads the user's hardware, selec…
Google launched agentic video understanding for its Gemini Flash models, enabling the AI to navigate video timelines instead of ingesting them at a fixed 1 frame per second, and reports up to 88% fewe…
NVIDIA released Personal AI Router (PAIR), an open source virtual inference router that distributes local AI requests across RTX, DGX Spark, and Mac nodes on a home network, proxying existing Ollama a…
Google DeepMind and Google Research released WeatherNext 3, an AI weather model that delivers global forecasts at up to 5 km resolution refreshed every hour, trained on raw weather station observation…
OpenAI released GPT-6 Astra, a computer-use model with a 1,050,000-token context window, available only to Trusted Access and Daybreak organizations due to its designation as the first model to reach …
Anthropic released anthropics/commerce-agents, an Apache-2.0 reference blueprint containing a shopping agent and a merchant agent for retail, travel, telecom, and entertainment, providing reusable sca…
Meta AI released Muse Spark 1.3, an agentic coding model that uses approximately 20% fewer tool calls and 25% fewer tokens than its predecessor Muse Spark 1.2. The model is designed to improve efficie…
Perplexity has open sourced Lily, a Rust and Metal inference engine for Qwen3.6-35B-A3B on Apple Silicon, achieving a 77.4% prefill speedup via fused grouped GEMM and an 89% gain from GPU-only routing…
Qwen Developers has open-sourced zg (zvec-grep), a local-first search layer that unifies ripgrep, BM25, and vector search behind a single interface, enabling agents to move from plain-language queries…
NVIDIA released Switchyard, a Rust proxy and library that routes and translates LLM traffic across OpenAI and Anthropic APIs, under Apache 2.0, but labels it pre-alpha and experimental, not for produc…
Google DeepMind released Gemini 3.8 Flash and Gemini 3.8 Flash Cyber on September 2, 2026, with both variants sharing the same core model but differing in safety mitigations. Gemini 3.8 Flash is gener…
Anthropic introduced Enterprise Frontier Safeguards (EFS), an architecture that gives regulated enterprise customers zero-data-retention privacy while enabling cross-session misuse detection by storin…
Meta Superintelligence Labs released Muse Voice Transcribe, a real-time audio perception model that performs streaming ASR, speaker diarization for 20+ speakers, and endpointing in a single autoregres…
Perplexity released Hybrid Compute for Mac, enabling its Perplexity Computer agent to split tasks between cloud frontier models and a local on-device model, with an on-device privacy gate deciding wha…
Anthropic released Claude Fable 5.1 and Claude Mythos 5.1, with Fable 5.1 scoring 52.6% on Terminal-Bench-Science 0.1 (up from 24.7% for Fable 5) and cutting cache reads by 75% to $0.25 per million to…
Researchers from Princeton University, Ant Group and Stanford University introduced AQuA, a two-part agentic framework for autonomous factor discovery and model development in quantitative finance, wh…
Gradium AI has released a new text-to-speech model, now the default across its API and Studio as of August 31, 2026, achieving an 81.0% human-rated pass rate on a 500-sentence hard-case set spanning f…
Keenable AI released NEEDLE, an open-source live search benchmark that regenerates its query set hourly from RSS feeds and Google Trends for news and daily from SEC XBRL, arXiv, Europe PMC, CourtListe…