Model releases, breakout repos, app launches, pricing shifts, security stories, and funding rounds from across the AI ecosystem. Each comes with our take on what it means for Vibe Builders and SMB owners. Updated hourly.
Open-OSS released privacy-filter on Hugging Face Hub, a token-classification model that detects personally identifiable information in text.
Google DeepMind launches Veo 3.1, its state-of-the-art video generation model, on Fal. Fal offers direct inference via HTTP API and web playground.
Fal.ai launches Kling Video 3 Image-to-Video with cinematic visuals, fluid motion, native audio, and custom elements. Access via HTTP API or web playground.
Lambda closed a $1 billion senior secured credit facility. The upsized deal builds on its August 2025 facility to expand AI factory footprint.
OpenRouter adds Tencent's free Hy3 preview, a high-efficiency MoE model for agentic workflows. It supports 262k context and configurable reasoning levels.
OpenRouter adds Alibaba's Qwen3.5 Plus multimodal model with 1M token context. It processes text, image, and video inputs at $0.40 per million input tokens and $2.40 per million output tokens.
Poolside's Laguna M.1 coding model is now free to use on OpenRouter, with a 128k context window and zero per-token cost. The Poolside coding-assistant product itself remains enterprise-only (on-prem or VPC deployment).
IBM Granite released granite-4.1-8b on Hugging Face Hub, an 8-billion-parameter text-generation model in the Granite 4 family.
Unsloth has released a local-runnable build of NVIDIA's Nemotron 3 Nano Omni on Hugging Face. The model is multimodal, sized so it fits on a consumer GPU, and is already at 48k downloads with 101 likes in its first days.
SulphurAI launched Sulphur-2-base on Hugging Face Hub, a text-to-video model running on the diffusers library. Supports download, fine-tuning, and inference for short-form generative video workflows.
OpenRouter launched a live test environment. The feature enables real-time testing of model routing across multiple AI providers.
Fal.ai launches a fast endpoint for FLUX.1 [dev] with LoRA support. It enables high-quality image generation using pre-trained LoRAs for personalization and styles.
Fal.ai launches FLUX.1.1 [pro], an upgraded FLUX.1 [pro] with superior composition, detail, and artistic fidelity. Access it via Fal's HTTP API or web playground.
Fal launches Kling Video v3 Pro image-to-video model with cinematic visuals, fluid motion, native audio generation, and custom element support. Access it via Fal HTTP API or web playground.
Fal releases FLUX.1 [schnell], a 12 billion parameter flow transformer that generates high-quality images from text in 1 to 4 steps. It supports personal and commercial use via HTTP API or web playground.
Z.ai releases GLM-OCR, a compact 0.9B multimodal OCR model, on Replicate as lucataco/glm-ocr. It tops OmniDocBench V1.5 at 94.62% with text, LaTeX formula, table parsing, and JSON schema modes.
Black Forest Labs releases flux-1.1-pro-ultra on Replicate. FLUX.1.1 [pro] supports ultra and raw modes with images up to 4 megapixels.
Black Forest Labs launches Flux.2 Flex on Replicate. The model enables maximum-quality image generation and editing with support for ten reference images.
xAI launches Custom Voices for Grok. It clones voices from one minute of speech atop recent Speech-to-Text and Text-to-Speech APIs.
lucataco releases sensenova-u1-8b-mot on Replicate. SenseNova U1 8B MoT enables unified multimodal tasks including native text-to-image generation.
Replicate publishes lucataco/gemma-4-31B-IT. Google's open-weight Gemma 4 31B Instruct VLM processes image and text inputs to generate text outputs.
Zed releases version 1.0 stable. The open-source code editor delivers high performance and multiplayer collaboration.
Build American AI, a nonprofit tied to a super PAC backed by OpenAI and Palantir executives, pays influencers to promote U.S. AI and warn of Chinese AI risks.
lucataco/z-anime launches on Replicate. This Z-Image Base fine-tune generates anime aesthetics from natural-language prompts with full negative prompt support and high-quality outputs for portraits, scenes, and…
Elon Musk testified that xAI trained Grok on OpenAI models. Model distillation draws scrutiny as labs block rivals from copying frontier AI.
