Skip to content
K = 0.731
All news · Today2 items

#Open source

All collected and published items, ordered by time. Picks are a selected subset.
Today10/17Sat
  1. vLLM · X77

    MOSS-Image goes open source with bilingual text rendering and experimental layout control

    MosiAI announced the open-source release of MOSS-Image, built on a 5B Diffusion Transformer (DiT), with Chinese and English text rendering, up to 4K-token prompts, and experimental JSON layout control using bounding boxes. Base and SFT checkpoints are available on Hugging Face, alongside open model weights and Apache-2.0 licensed inference code. vLLM congratulated the team and announced Day-0 support through vLLM-Omni.

    Why it matters: The release provides a concrete example of combining bilingual text rendering with experimental bounding box layout control, giving developers a reference for image generation workflows that require explicit text and object placement.

10/16Fri
  1. vLLM · X32

    vLLM Supports Muse Spark Open in a Story Highlighting Its Capabilities

    @90VuHwwuzNhhW3h says vLLM supports Muse Spark Open and made a little film to tell its story, using open-source models served by vLLM-Omni. The inference engine also gets a role behind the camera. https://t.co/n3gvdWTC6T The main author calls the film cute and says it brings multimodal understanding, 1M context, agentic planning, DSpark, and cross-platform support to life. https://t.co/yHPXC7qfdM

  2. Aravind Srinivas · X71

    Perplexity open-sources SPACE, the sandbox platform powering Perplexity Computer

    While sandboxes are a commodity, serving millions of them under production-grade traffic requires attention to detail to scale agents for businesses and consumers. Perplexity is open-sourcing SPACE, the sandbox platform that powers all of Perplexity Computer, to support the ecosystem. https://github.com/perplexityai/space

    Why it matters: The distinction between commodity sandboxes and operating millions under production traffic gives context for the engineering demands of scaling agents for businesses and consumers.

  3. Perplexity · X68

    Perplexity open sources SPACE, the sandbox platform behind Perplexity Computer

    We're open sourcing SPACE, the sandbox platform that runs every task in Perplexity Computer. SPACE gives each agent task its own Firecracker microVM and controls credentials and outbound traffic from outside the VM. Available now under Apache 2.0: https://t.co/DzSmq6o1xA

    The same story is shown in the picks as 《Perplexity open-sources SPACE, the sandbox platform powering Perplexity Computer》

10/15Thu
  1. 量子位82

    Skywork AI releases open-source Matrix-Game 4.0 for real-time gaming on one RTX 5090

    Skywork AI released Matrix-Game 4.0, an open-source game world model that reportedly runs at 480p and 60 FPS on one RTX 5090, with input response under 0.15 seconds and peak VRAM usage of approximately 18.5GiB. A game engine maintains state and executes rules, while a video model generates visuals from structural frames and dynamic captions. Model weights and code are available through https://huggingface.co/Skywork/Matrix-Game-4.0 and https://github.com/SkyworkAI/Matrix-Game.

    The same story is shown in the picks as 《Kunlun Tech releases and open-sources Matrix-Game 4.0 for single-GPU real-time gameplay》

  2. 宝玉 · X34

    How BaoCut Aligns English and Chinese in Bilingual Subtitles

    The author introduces an episode about subtitle translation challenges and how their open-source project BaoCut aligns English and Chinese when the languages use different word orders. They describe subtitles appearing before someone speaks, sentences split into pieces that do not match the spoken translation, and two or three lines covering half the screen. The post links to the project: https://t.co/c3VnQEt8x3

10/14Wed
  1. 量子位60

    APUS AI Open-Sources Reflex Agent With Model-Driven Control and Persistent Memory

    APUS AI announced its MIT-licensed Personal Agent, Reflex Agent, on October 13, offering desktop, Web, and SDK access. Its Reflex decision model handles routing, skill selection, result review, and execution supervision, while Continuum retrieves historical information to support long-running tasks without exceeding individual model calls’ input limits. In material supplied by APUS AI and republished by QbitAI, the company reports responses up to 3 times as fast as Claude Code and Codex in tested scenarios; the agent is at https://github.com/APUS-AI-Lab/APUS-Reflex-Agent and the decision model at https://huggingface.co/apus-ailab/APUS-Reflex.

10/13Tue
  1. Simon Willison36

    Simon Willison Builds web-edit 0.1a0 to Edit Files From a Phone

    Simon Willison built web-edit 0.1a0, a Python web app for viewing and editing files in a folder, after wanting to edit a credentials file from his phone without sending those credentials through the LLM. He intends to run it over Tailscale so an agent can start the app and he can edit files directly from his phone. The editor uses CodeMirror 6; Willison said he had considered Monaco but found its mobile experience was not yet great.

10/12Mon
  1. Qwen · X61

    Qwen introduces QwenGyre for elastic RL in extreme-long-horizon agent training

    Qwen introduces QwenGyre, bringing elastic RL to extreme-long-horizon (XLong) agent training. Its Elastic Scheduler reallocates GPUs between rollout and training without interrupting live executions, while its Trajectory Processor converts branching executions into prefix-shared trajectory trees with bounded, deduplicated training samples. Scaled to Qwen3.8-2.4T-A95B with ~700K tokens per rollout, QwenGyre raises NL2RepoBench from 52.5% to 58.5% in 48 steps and cuts end-to-end training time by ~44% vs Async. Paper: https://arxiv.org/abs/2609.33848

10/11Sun
  1. Clément Delangue · X53

    OpenAI’s math problems become open-source RL environments on Hugging Face

    We turned @OpenAI's math problems into open-source RL environments on @huggingface! They are early and rough, and the verifier only accepts the exact original formalization, so an equivalent proof can still score 0. I find this an exciting direction for turning open research into executable environments anyone can build on to train better open models! https://huggingface.co/datasets/FineEnvs/openai-math

  2. 量子位62

    Unofficial DeepSeek-Bot plugin adds multi-bot teams to DeepSeek Harness

    An unofficial plugin called DeepSeek-Bot builds a multi-bot team interface on top of DeepSeek Harness, letting each bot have its own name, role, model and persistent long-term memory. A Main Bot can manage other bots and group chats to cut token use, and the project uses an everything-is-a-plugin design under Apache License 2.0. It installs on DSH desktop and web, where Node.js 22.19 or higher and pnpm are required.

10/10Sat
  1. 雷锋网42

    Columbia's FlexiTac Open-Source Tactile Sensor Platform Hits 1.5 Years

    Columbia University's Li Yunzhu team presented progress on FlexiTac, an open-source flexible piezoresistive tactile sensor platform, at an IROS 2026 workshop. The sensor assembles in 3 minutes, ships with a full BOM and readout software, and integrates with the LeRobot robot-learning framework. Analog Devices is developing a commercial version with tactile resolution five times that of a human finger, produced via semiconductor manufacturing.

10/9Fri
  1. vLLM · X40

    vLLM Semantic Router team ships Decision 2.0 for single-pass routing and classification

    Great work from the vLLM Semantic Router team on Decision 2.0. Multiple questions about one input, answered in a single forward pass with per-option probabilities. Useful for routing and classification. Quoted post by @XunzhuoLiu: In every size. Decision 2.0: six open decision models, 0.6B to 27B. On the #Jev Decision Index 0.3, each one tops the same-size open models we lined it up against. #Trending collections on Hugging Face 🤗 Decision 2.0: https://huggingface.co/collections/vllm-sr/decision-20 Vela 2.0: https://huggingface.co/collections/vllm-sr/vela-20

10/8Thu
  1. Hugging Face Blog40

    TII launches Falcon-ASR, a 1.6B-parameter Arabic speech recognition model focused on Emirati dialects

    Abu Dhabi's Technology Innovation Institute (TII) released Falcon-ASR, a 1.6B-parameter speech recognition model focused on Arabic and Emirati dialects, with support for English, French, Spanish, and Portuguese as well. Its average WER across six Arabic test sets is 20.92%, 2.25 percentage points below the best public result of 23.17% in the leaderboard snapshot used. In an internal Emirati dialect evaluation, it achieved a 22.73% WER and 10.19% CER, outperforming Qwen3-Omni.

  2. vLLM · X62

    vLLM v0.31.0 ships with DeepSeek-V4.1-Flash support and fast restart

    vLLM v0.31.0 is live, spanning 717 commits from 307 contributors, including 96 first-timers. Highlights include DeepSeek-V4.1-Flash support with FlashMLA mega attention and NVFP4 KV caching, a vllm preload fast-restart path that keeps weights in GPU memory, Model Runner V2 speculative decoding, and large-scale serving features such as MoonEP balanced all-to-all and DeepEPv2. Full release notes are on GitHub.

  3. 雷锋网44

    Galaxy General, Tsinghua, and collaborators propose LATENT, teaching humanoids tennis from imperfect human data

    The GaLbot team at Galaxy General, together with Tsinghua, Peking University, and others, proposed LATENT, which trains humanoid robots to play tennis using about five hours of imperfect motion data from five amateur players. The method first trains a motion tracker and compresses it into a latent action space with a Latent Action Barrier constraint. On a Unitree G1, it achieved a 90.90% success rate on real forehands and 77.78% on backhands.

  4. vLLM · X62

    vLLM-Omni technical report proposes a unified serving runtime for omni-modality generation

    The vLLM project shared the vLLM-Omni technical report, describing a unified serving runtime for omni-modality generation. It argues that speech assistants, visual generation, world models and robot loops diverge from a single text decode loop into multi-stage autoregressive pipelines, iterative diffusion and stateful sessions, which LLM servers and diffusion stacks each cover only partially. vLLM-Omni acts as a shared control plane with an orchestrator, specialized engines, a connector and one session path. Paper and repo links are included.

  5. Perplexity · X58

    Perplexity releases pplx-embed-v2-late multimodal embedding models

    Perplexity is releasing pplx-embed-v2-late, described as two late-interaction embedding models that retrieve text, images, and pages through a shared embedding space for cross-modal querying. The company says both models reach frontier performance and are publicly available on Hugging Face, with a blog post at perplexity.ai/hub/blog/multimodal-embeddings-beyond-a-single-vector.

10/7Wed
  2. Simon Willison40

    EmbeddingGemma 2 adopts the Apache 2.0 license

    EmbeddingGemma 2 uses the Apache 2.0 license. Simon Willison argues that embedding models should not be locked into proprietary hosted services. They often require computing and storing thousands or millions of vectors; if a provider discontinues a model, users may have to pay to recompute them. He would rather pay for a hosted version while retaining the option to run open weights or switch providers if the service shuts down.

10/6Tue
  1. Hugging Face Blog26

    Falcon-Emirati-7B: teaching LLMs the Emirati dialect, culture, and feel for language

    Falcon-Emirati-7B is a specialized model for Emirati dialects built on Falcon-H1-Arabic. It aims to understand and generate Emirati Arabic like a native speaker. It uses the Falcon-H1 hybrid architecture, running Mamba alongside Transformer attention. The base model family comes in 3B, 7B, and 34B sizes, with a context window of up to 256K tokens.

10/5Mon
10/4Sun
10/3Sat
  1. Hugging Face Blog62

    Ai2 open-sources AstaBrief 8B, a scientific report-generation model

    Ai2 open-sourced AstaBrief 8B, a model that turns research questions and retrieved literature excerpts into reports with citations. It was trained from Qwen3-8B using SFT and DPO and is now available as Asta's Fast mode.

    Why it matters: The source describes AstaBrief's training-data construction and citation-filtering details, which can be adapted to deploy other long-form generation models.

10/2Fri
  1. Simon Willison32

    pwasm 0.2a0 released: Claude Opus 5.5 improves pure-Python WebAssembly engine

    Simon Willison's pure-Python WebAssembly engine pwasm reached version 0.2a0 after Claude Opus 5.5 took over and made 42 commits. It now supports almost the entire WASM specification. The PyPI wheel bundles working WASM builds of MicroPython, QuickJS, and Micro QuickJS. This release is still alpha, and the author says it is completely untrusted.

10/1Thu
  1. TechCrunch · AI42

    Photon raises $4.5 million seed round after holding a funeral for mobile apps, betting on agents instead

    Photon raised a $4.5 million seed round co-led by Gradient and A*, with participation from Vercel, Sequoia (HongShan), and others. The startup helps developers build agents for iMessage and WhatsApp and says it has attracted more than 40,000 registered developers. Revenue grew 10-fold in four months, and message volume grew fivefold last month. Its managed platform offers 99.95% availability and is certified for SOC 2 Type II and HIPAA compliance.

9/30Wed
9/25Fri
9/24Thu
9/21Mon
  1. Microsoft Research40

    Microsoft open-sources RetroChimera retrosynthesis model in a paper published in Nature

    Microsoft has published RetroChimera, a retrosynthesis-prediction framework, in Nature and open-sourced its implementation and weights. It combines the Transformer model R-SMILES 2 with the GNN model NeuralLoc, using a learned ensemble to rerank their predictions and performing strongly on both common and rare reaction types. In blind tests, chemists preferred its single-step reaction predictions. It found complete routes for nine of ten challenging targets, compared with five for a de novo model, four for an editing model, and two for NeuralSym.

  2. Hugging Face Blog48

    Hugging Face releases tokenizers v1: benchmarks show major encoding and decoding speedups

    Hugging Face has released tokenizers v1, producing exactly the same token IDs as v0.23 while running up to dozens of times faster. The project splits a single crate into a workspace and introduces bitcannon, which replaces regex tokenization with SIMD bit operations, along with an allocation-free merge loop, thread-local word cache, and native multithreaded parallelism.

9/20Sun
9/19Sat
9/3Thu