AI Briefing#open-source
The latest AI and developer news about #open-source, with the original source and a short summary.
The latest AI and developer news about #open-source, with the original source and a short summary.
RelateAnything supports an inference speed of 17.2ms per image on the RTX 3080 Laptop GPU using TensorRT FP32 and has released the RA-4M training corpus, which includes over 474,000 images and 4.3 million relationships.
The MIT-licensed tool allows users to monitor and respond to permission prompts from Claude Code, Codex CLI, and OpenCode across multiple folders via desktop or phone.
Users must now install the 'modelscope-hub' package to access ModelScope CLI tools, as the main 'modelscope' package no longer includes executables.
E2B Embed is licensed under Apache-2.0 and requires no E2B account or license key.
Perplexity advocates for layered security architectures to contain AI agent failures, citing recent incidents involving OpenAI and Anthropic models.
The new leaderboard uses objective metrics like WER and RTFx to evaluate open-source TTS models, reducing assessment time from weeks to hours.
A local-based multi-agent management tool that solves the issue of restoring agent configurations taking 30 minutes upon reboot.
Jeeves achieves 0.935 on JevBench overall, outperforming the baseline Jev model's 0.866 by integrating reasoning chains and a block-4 diffusion drafter.
Kumo Tabular achieves state-of-the-art accuracy on TabArena with ELO 1950 while being 17x faster than LimiX-2, trained entirely on synthetic data.
The project introduces webbrain-vl-2-450M, a specialized vision model trained on browser screenshots that runs locally via WebGPU.
The open-source vllm-xtu-moe project achieves up to ~20× faster prefill than CPU-only paths for medium contexts by streaming expert weights from CPU to GPU.
The tool supports over 40 agents via the Agent Client Protocol and has reached approximately 50,000 downloads.
The new authorization brings Nvidia's total remaining buyback capacity to $235 billion, the largest in history.
The open-source multi-agent harness Munder Difflin v0.5.4 introduces a local transcription model that serves as an alternative to Whisper Flow and Granola.
A Jev-compatible open-weight model that returns probability distributions in a single forward pass to address LLM hallucination and latency issues.
The release includes a new git add --resolved command to prevent staging unrelated changes and optimizations that reduced merge-base traversal time from 0.68 seconds to 0.01 seconds in large monorepos.
Vinext 1.0 achieves over 99% test compatibility for key features and introduces cache warming to move page prerendering from build time to Cloudflare's network.
Cloudflare has released EmDash 1.0, a stable, open-source CMS built on Astro that features a decentralized plugin registry based on AT Protocol and sandboxed plugin execution.
Forge is released under the Apache 2.0 license and currently powers the cf CLI, with plans to generate Cloudflare's API documentation and SDKs in the coming months.
Vite+ 1.0 unifies the toolchain with Vite 8, Vitest 5, and Rolldown, while new Rust-based tools offer up to 18x speedups over competitors.
CLM-v0.1-8B, based on Qwen3-8B, released with candidate action selection speed up to 13x faster than Jev
The MIT-licensed curriculum now includes six downloadable volumes and supports eight languages.
The MIT-licensed tool enables AI agents like Cursor and Claude Code to automate tasks using the user's existing Chrome or Edge sessions via a dedicated CLI and browser extension.
The platform combines the open-source NVIDIA OpenShell runtime and the NVIDIA Sentry reference system design to enable full-stack governance across software, hardware, compute, and robotics systems.