Chinese researchers just dropped ASI-Evolve, a 100% open-source agent framework designed to conduct AI research autonomously. It reads papers, designs experiments, runs code, analyzes results, and iteratively improves itself through a closed-loop cycle. Both code and paper are out.
Colleague-Skill is a viral open-source project that digests chat logs from WeChat or Lark to clone departed teammates into working AI agents. By decoupling personality from task execution, it faithfully mimics their tone, coding rules, and communication habits. It already racked up over 6.7k GitHub stars in three days.
H Company just launched Holo3, a suite of computer-use models designed to automate complex desktop workflows. The flagship Holo3-122B-A10B tops the OSWorld benchmark with a 78.85% score using only 10B active parameters, while the smaller 35B version is open-sourced under an Apache 2.0 license.
Researchers just open-sourced Gen-Searcher, a search-augmented agent framework that scours the web for reference images and facts before generating visuals. By grounding diffusion models in real-time retrieval, it tackles visual hallucinations on obscure entities across both open and closed models.
OpenMontage is an open-source AI video production system that hooks directly into assistants like Claude Code and Cursor. Give it a prompt, and it handles research, scriptwriting, voiceover, and editing. It even shows you storyboards and cost estimates first, clocking a 60-second sample animation at just $1.33.
Developer Yatharth Sharma just open-sourced LuxTTS, an ultra-fast voice cloning model that requires only a 3-second audio sample. Fitting in under 1GB of VRAM and outputting 48kHz audio, it reaches 150x real-time speed on GPUs and outpaces real-time even on basic CPUs. High-fidelity voice cloning can now run on almost any budget laptop.
Midjourney engineer Cheng Lou just launched Pretext, an open-source, 15KB TypeScript library that fixes a persistent web performance bottleneck. Built using AI vibe coding tools like OpenAI's Codex and Anthropic's Claude, it handles multiline text layout entirely in userland. That bypasses costly DOM reflows and lets text wrap around moving objects in real time.
A new open-source agent skill lets developers recreate any website with pixel-perfect accuracy using /clone-website. Instead of guessing layouts in one massive prompt, it breaks components into strict specs under 150 lines and builds them in parallel git worktrees before running visual diffs. It basically fixes the habit of AI faking CSS spacing and fonts.
Meta just dropped SAM 3.1, adding a shared-memory approach called Object Multiplex to its visual segmentation model. It delivers roughly 7x faster inference when tracking 128 objects on a single H100 without dropping accuracy, beating prior scores on 6 out of 7 video benchmarks. The model checkpoints are available to request on Hugging Face.
Babylon.js just launched version 9.0, calling it the framework's biggest update yet. The release introduces Clustered Lighting to render thousands of lights smoothly across WebGPU and WebGL 2, alongside a Frame Graph pipeline that cuts GPU memory by up to 40%. It also rolls out a Node Particle Editor and expands Gaussian Splat support built with Adobe.
Real Restorer just launched as a specialized open-source model designed to rescue ruined photos, from wiping out glare and noise to restoring scratched black-and-white shots. It matches top closed-source models in benchmarks, but you will need a beefy GPU to host its 42 GB weight package locally.
Cohere just dropped Transcribe, its first open-source speech recognition model designed for note-taking and voice analysis. Weighing in at 2 billion parameters, it runs on consumer GPUs, processes 525 minutes of audio per minute, and supports 14 languages. It topped the Hugging Face Open ASR leaderboard with a 5.42 WER, and Cohere is also offering free API access.
ByteDance just open-sourced Lark CLI under the MIT license, specifically designed for human developers and AI agents. It spans 18 enterprise domains—including Docs, Base, and Calendar—with over 200 curated commands. Why it matters: popular AI coding tools like Claude Code can now hook into and operate your entire Lark workspace with zero setup.
A new 15B open-source video generator called DaVinci Magihuman just dropped, capable of natively generating video with audio across multiple languages. In self-reported blind tests, it beat LTX 2.3 60% of the time with zero censorship, though running even its distilled version requires a hefty 61GB download.
Alibaba just open-sourced PrismAudio, a lightweight 518M-parameter model that generates synchronized sound effects for silent videos. By combining reinforcement learning with chain-of-thought reasoning, it generates 9 seconds of audio in just 0.63 seconds, outperforming multi-billion-parameter rivals in both speed and benchmarks.
Ai2 just released MolmoWeb, an open-weight visual web agent that operates purely from browser screenshots without parsing HTML or accessibility trees. Available in 4B and 8B sizes, it arrives with MolmoWebMix, a dataset of 30,000 human web trajectories, beating open-weight rivals on live benchmarks like WebVoyager.
ByteDance just open-sourced DeerFlow 2.0, an MIT-licensed agent orchestration harness designed for complex, multi-hour jobs. Instead of acting as a thin wrapper, a lead agent breaks tasks down and spins up isolated sub-agents inside Docker containers with dedicated filesystems and persistent memory. It works with local or cloud models like DeepSeek and GPT-4.
Pascal Editor just launched as a 100% free and open-source 3D architecture editor running entirely in the browser. Powered by WebGPU and React Three Fiber, it supports real-time editing of walls, slabs, and zones with GPU-accelerated rendering. It brings desktop-grade architectural modeling to the web without the typical enterprise software price tag.
GreenSock just rolled out gsap-skills, an official skill pack that teaches 40+ AI agents—including Cursor, Claude Code, and Copilot—how to write clean GSAP code. It covers timelines, ScrollTrigger, and React patterns, basically stopping your coding assistant from hallucinating outdated animation syntax.
MetaClaw just dropped an open-source continual meta-learning framework that turns regular conversations into training signals for LLM agents. It synthesizes reusable skills immediately after failures, then triggers cloud LoRA updates while you sleep or take meetings. That gives personal agents a way to self-improve over time with zero downtime and no local GPU required.
Unsloth just launched Unsloth Studio, an open-source local UI that turns PDFs or CSVs into datasets and fine-tunes models 2x faster with 70% less VRAM. It packages visual dataset creation, training, and GGUF export for 500+ open models into one interface, and is available now on GitHub.
SegviGen just open-sourced a 3D part segmentation framework built on TRELLIS.2 that splits meshes into individual components via single clicks or 2D guide images. It outperforms previous SOTA models by 40% in interactive segmentation while requiring far less supervision, though local inference demands at least 24GB of VRAM.
Google just open-sourced Spark VSSR, a powerful video super-resolution framework capable of sharpening degraded clips and restoring old films. The release includes inference models, full training code, and datasets, though running the 42.2GB package demands a high-end GPU. Basically, creators get state-of-the-art video upscaling that they can run and fine-tune locally.
Google just open-sourced the Colab MCP Server, enabling any MCP-compatible AI agent to programmatically control Google Colab notebooks straight from your terminal. Tools like Claude Code or Gemini CLI can now create cells, run Python scripts, and render visualizations live in the cloud. It basically turns Colab into an automated sandbox so agents no longer run risky, resource-heavy code on your local machine.
Researchers Albert Gu and Tri Dao just released Mamba-3, an open-source state space model architecture under the Apache 2.0 license designed to outpace Transformers. Its MIMO variant hits a 57.6% average benchmark accuracy—a roughly 4% relative bump over Transformers—while matching Mamba-2's quality using half the state size. By introducing complex-valued states and cutting GPU idle time, it targets reasoning gaps and slashes inference latency.
Mistral just unveiled Mistral Small 4, consolidating reasoning, multimodal vision, and agentic coding into a unified 119B MoE model with only 6B active parameters. It offers a 256k context window, a toggleable reasoning dial, and cuts completion latency by 40% compared to version 3. Best of all, it's released under Apache 2.0 and beats competing models with far fewer output tokens.
MiroMind just launched MiroThinker-1.7 and MiroThinker-H1, open-source heavy-duty research agents that bake real-time verification directly into their reasoning loops. Packing 235B parameters alongside a 61GB mini variant, H1 tops demanding benchmarks like BrowseComp by auditing intermediate steps and proof trails before serving answers.
Alibaba just open-sourced Fun-CineForge, a multimodal dubbing model that syncs speech directly to characters' lip movements and timestamps. By treating time as an explicit modality, it handles fast speaker switches and facial occlusions cleanly, turning automated film dubbing into a realistic studio workflow.
Tsinghua University researchers just open-sourced OpenMAIC, an AI platform that turns any topic or document into a full interactive classroom. Instead of watching static video lectures, users learn alongside AI teachers and classmates who speak, draw formulas on whiteboards, and engage in real-time debates. It basically shifts online learning into an active virtual classroom under a permissive MIT license.
NVIDIA just launched the Nemotron Coalition, teaming up with Mistral AI, Cursor, Perplexity, LangChain, and other top AI labs to build open frontier foundation models. Their first milestone is a joint base model co-developed with Mistral AI on DGX Cloud, which will be open-sourced to underpin the upcoming Nemotron 4 model family.
Tencent just open-sourced Covo-Audio-Chat, a 7B end-to-end audio language model. It processes continuous audio directly in a unified architecture to enable native, low-latency full-duplex conversations that can listen, speak, and handle interruptions on the fly. Weights and code are now available on Hugging Face and GitHub.
Texas A&M researchers just open-sourced Pulse of Motion, a lightweight AI designed to fix "chronometric hallucinations" in generated videos. Instead of guessing playback speed, the sub-2GB model calculates the true physical frame rate directly from visual dynamics and retimes the clip so movements actually obey gravity.
