
TIPOv2 Rebuilds Itself as a Sparse AI to Write Your Image Prompts
TIPOv2 rebuilds the popular prompt-expansion model as a 991M sparse MoE with 193M active parameters, five datasets, and a new training stack.
20 stories

TIPOv2 rebuilds the popular prompt-expansion model as a 991M sparse MoE with 193M active parameters, five datasets, and a new training stack.

NVIDIA's new 30B mixture-of-experts model runs 3B active parameters, hits 1M context, and targets the boring execution layer of always-on agents.

Pika launches a $10/month API Club giving builders access to 100+ gen media models at up to 88% off aggregator prices through a single API

Artificial Analysis launches the Endpoint Accuracy Index, revealing how much model intelligence providers silently trade away for speed and cost savings

Black Forest Labs launches FLUX 3 Video with native audio, up to 20-second clips, a draft-then-enhance workflow, and pricing that undercuts most competitors.

Mistral's 3B open-weights safety classifier beats models 7x its size by turning content moderation into a plain-English yes/no question
How Moonshot fit a 2.8 trillion parameter model into real hardware without breaking it.

Google Cloud API Gateway now routes traffic across Gemini, Claude, and OSS models from a single OpenAI-compatible endpoint, no proxy servers required.

Goodfire's Silico goes public: an AI agent that autonomously runs interpretability experiments at 2.8 trillion parameter scale, for $1,000/month

Cursor open-sources Mixture-of-Kittens, a production MoE megakernel that delivers 2.37x faster throughput and 1.41x end-to-end training speedup over DeepEP on GB300 NVL72s.

Bolt.new launches a free Template Marketplace of designer-built, full-stack apps you can clone and customize with a prompt

Liquid AI's LFM2.5-2.6B runs full agentic workflows on-device, beating models nearly 4x its size on tool use and instruction following benchmarks.

PRISM2 is a 4.4B-parameter pathology foundation model that matches specialized cancer detectors using plain question-answering, with weights freely available for research.

A community repo bundles five ready-to-run ComfyUI workflows for MiniMax H3, unlocking Director-style camera control for local video generation, editing, and character swaps.
Tencent's new ASR model plugs directly into Hy3's language brain, hitting ~3% WER across Mandarin, English, and Cantonese while understanding context, not just words.
Microsoft Research drops SocialRL, EvoLib, and PazaBench V2 — five releases that push the frontier of agent negotiation, continual learning, and African speech AI.

webAI's 3B formal-logic model matches an OpenAI 120B on four of five reasoning lanes and runs on a Raspberry Pi

OpenAI's GPT-Live gets a full engineering teardown: WARP protocol, stateful inference, and async delegation make voice feel truly live

Microsoft Research open-sources Orchard, a Kubernetes-native framework that trains competitive coding, browser, and productivity agents using small models and reusable infrastructure

v0's new API v2 lets you embed its full app-building agent — prompt, preview, iterate, deploy — directly into your own products and pipelines