World Wire

Long-horizon agents have turned LLM serving into an input-heavy workload. Repeated prefills and million-token contexts leave KV caches that strain HBM, SSD capacity, and bandwidth. DeepSeek AI built its newest release around that exact bottleneck. DeepSeek-V4.1-Flash is a multimodal Mixture-of-Experts model with 552B backbone parameters, 196B additional Engram parameters, and a 1M-token context window. It activates 8B parameters per token during prefill and 16B during decode. The main number is a global KV cache footprint of 890 bytes per token, about 1/4 of DeepSeek-V4-Flash and roughly 437x smaller than DeepSeek-V1. Is it deployable? Yes. Open weights ship under an MIT license with vLLM, SGLang, and Transformers paths on Hugging Face, and the research team describes a public API with low, high, and max reasoning tiers. Causal Encoder-Decoder: Half the Prefill The 40-layer backbone is split into a 20-layer causal encoder and a 20-layer decoder. Inspired by YOCO , the ...

LandingAI has shipped Agentic Document Extraction (ADE) Gen2 , a rebuild of its document intelligence stack around a new model family called DPT-3. Gen1 treated a document as a flat list of chunks. Gen2 treats it as a tree, prices it by the characters it returns rather than by the page, and grounds every answer back to a specific line or word on the page. LandingAI team frames the release around three themes: affordability, agent ready outputs, and atomic grounding. Is it deployable? Yes. ADE Gen2 is generally available now. Developers start free in the ADE playground . Enterprises can run it in US or EU cloud, in their own VPC on AWS, Azure or Google Cloud, inside Snowflake, or on premises including air gapped environments. Two parsing models instead of one Gen2 splits parsing into two models so the workload picks the price. DPT-3 Verity transcribes digitally created documents deterministically and returns a bounding box and a confidence score for every word. It targets...

Google has open-sourced Mantis , a stack-agnostic toolkit of security review skills that lets an AI coding agent run the whole vulnerability lifecycle. It finds a suspected flaw, strips the false positives, reproduces the bug inside a sandbox, writes a minimal patch, re-attacks that patch, and scores the residual risk. Mantis is not a scanner you aim at a repository and walk away from. It is a set of slash commands your existing coding agent loads, plus a strict set of rules about where that agent is allowed to execute code. Is it deployable? Yes for local and internal evaluation, not yet for production. You can clone it today and run it with Gemini CLI , Antigravity CLI, the Google ADK, or any comparable agent framework. The pipeline Mantis publishes each stage as a separate skill directory, invoked as a slash command and chained sequentially. A supervisor skill, /mantis-meta-agent , can drive the whole loop in a long-lived session. The early stages learn the tar...

Voice agent teams keep hitting the same wall. The catalog holds 400 voices and the brief asks for the one that is not in it: a Quebecoise receptionist for a Montreal dealership, a narrator in his sixties with lecture hall authority. Briefs outnumber any catalog, and cloning closes the gap one speaker at a time, each carrying sourcing, consent and a licence. Gradium , the Paris-based voice AI company spun out of the Kyutai research lab, has shipped a different answer. Voice Design reads a written description and returns complete new voices in a few seconds. No reference audio, no speaker, no rights to clear. Is it deployable? Yes, Voice Design is live in the Gradium API and in Studio , free on every plan including the free tier, and a kept voice runs on the same streaming Text-to-Speech endpoint as any catalog voice, at the same latency and output formats. The casting brief is the API The description is the only input the model gets. Gradium’s documentation list...