# Open Source — AI updates

Open models, tools, and releases — and the shift from renting AI toward owning and self-hosting it.

- [Vercel open-sources durable-harness for crash-resilient coding agents](https://www.parallelquant.com/posts/vercel-open-sources-durable-harness-for-crash-resilient-coding-agents-67a27b) (2026-10-21, MarkTechPost): Vercel released durable-harness, an open-source package that logs every model and tool call. This lets sessions of Pi, OpenCode, Claude Code and Codex resume after crashes, timeouts and redeploys. It runs on Vercel, Postgres or a laptop.
- [Recursive releases Flint, open-weight 40B coding model for laptops](https://www.parallelquant.com/posts/recursive-releases-flint-open-weight-40b-coding-model-for-laptops-06bd6f) (2026-10-21, MarkTechPost): Recursive released Flint, an Apache 2.0 mixture-of-experts coding model with 40B total and 2.5B active parameters. It reports 74.2% on SWE-bench Verified and 125+ tokens per second on an M4 Pro MacBook Pro. TechCrunch reports it was built largely by Recursive's own AI research system.
- [AMD and Intel to share one PyTorch CPU inference path](https://www.parallelquant.com/posts/amd-and-intel-to-share-one-pytorch-cpu-inference-path-9c39fb) (2026-10-21, Tom's Hardware): AMD and Intel are collaborating to bring PyTorch CPU inference optimizations under the unified oneDNN library. The goal is a single code path for AI inference on CPUs from both vendors.
- [Maintainers say a quarter of AI-era pull requests are slop](https://www.parallelquant.com/posts/maintainers-say-a-quarter-of-ai-era-pull-requests-are-slop-59b288) (2026-10-21, The Register): Open source maintainers estimate about a quarter of incoming pull requests are low-quality, rising to 31% on larger projects. The report says AI cuts coding time while also flooding project inboxes with poor submissions.
- [DeepSeek open-sources DeepInfer, the engine behind V4.1-Flash](https://www.parallelquant.com/posts/deepseek-open-sources-deepinfer-the-engine-behind-v4-1-flash-a67e88) (2026-10-20, The Decoder): DeepSeek released DeepInfer, the inference framework that serves its fast, cheap V4.1-Flash model. It is built on vLLM and aims to cut compute costs and energy use when running models.
- [Hackers breach OpenCode platform, steal API keys and run up charges](https://www.parallelquant.com/posts/hackers-breach-opencode-platform-steal-api-keys-and-run-up-charges-17b6dc) (2026-10-19, TechCrunch): Hackers got into part of the open-source AI coding agent OpenCode's online platform and stole some users' API keys, running up charges on their AI accounts. The AI tools the team turned to for help kept refusing to inspect the malicious code.
- [Google open-sources Harness2, which rewrites agent harnesses at test time](https://www.parallelquant.com/posts/google-open-sources-harness2-which-rewrites-agent-harnesses-at-test-time-1ab254) (2026-10-18, MarkTechPost): Google Research released Harness2 under Apache-2.0, a framework that rewrites an agent's harness at test time. One refinement step lifted GLM-5 by 11.5 points on WorkBuddy-Bench with no weight updates.
- [Kyutai releases REKI, a 2.4B robot model with efficient memory](https://www.parallelquant.com/posts/kyutai-releases-reki-a-2-4b-robot-model-with-efficient-memory-2131b0) (2026-10-17, MarkTechPost): Kyutai, with Inria and École des Ponts, released REKI, a 2.4B-parameter vision-language-action (VLA) model for robot manipulation. It is pretrained only on public data and keeps a cheap memory of its recent past. On three out-of-distribution benchmarks, only models about twice its size beat it, and all weights are public.
- [Perplexity open-sources SPACE, its Firecracker microVM sandbox for agents](https://www.parallelquant.com/posts/perplexity-open-sources-space-its-firecracker-microvm-sandbox-for-agents-aa2d2d) (2026-10-17, MarkTechPost): Perplexity open-sourced SPACE, the sandbox platform behind each task in Perplexity Computer, under Apache 2.0. Each agent task runs in its own Firecracker microVM, with credentials and outbound traffic controlled from outside the VM.
- [StepFun releases Step 5 Preview, a 600B open-weight MoE model](https://www.parallelquant.com/posts/stepfun-releases-step-5-preview-a-600b-open-weight-moe-model-6d2773) (2026-10-17, MarkTechPost): StepFun released Step 5 Preview as open weights, with FP8 and BF16 checkpoints on Hugging Face and ModelScope. It is a sparse Mixture-of-Experts model with 600B total and 27B active parameters per token, and supports up to 1M tokens of context. It is built for long-horizon agentic work.
- [DeepSeek releases V4.1-Flash-1017 and opens Search and Fetch APIs](https://www.parallelquant.com/posts/deepseek-releases-v4-1-flash-1017-and-opens-search-and-fetch-apis-39b027) (2026-10-17, The Decoder): DeepSeek released V4.1-Flash-1017 as an open-weight model that it says beats larger rivals on agentic tasks, finishes faster and writes shorter outputs than the September version. It also opened its internal search engine through Search and Fetch APIs at $0.75 per 1,000 calls. API users in tools like Claude Code and Codex get free access for now.
- [China plans export controls on open-weight models and 23 AI technologies](https://www.parallelquant.com/posts/china-plans-export-controls-on-open-weight-models-and-23-ai-technologies-38d096) (2026-10-16, The Decoder): China plans to restrict exports of 23 key AI technologies, including model weights, source code and training methods for systems at 10^26 FLOPs or 2 trillion parameters. A fast-track rule would still clear open-source releases once they are registered.
- [Linus Torvalds posts first LLM-assisted patch to Linux kernel](https://www.parallelquant.com/posts/linus-torvalds-posts-first-llm-assisted-patch-to-linux-kernel-6e685a) (2026-10-16, Tom's Hardware): Torvalds submitted his first patch made with AI tooling to the Linux kernel. A synthetic test reportedly showed a 23x performance improvement. The move follows his earlier directives about how AI tools may be used in kernel work.
- [Black Forest Labs releases FLUX 3 Image Klein, runs in 8 GB VRAM](https://www.parallelquant.com/posts/black-forest-labs-releases-flux-3-image-klein-runs-in-8-gb-vram-c4cae3) (2026-10-15, The Decoder): Black Forest Labs released FLUX 3 Image Klein, a 5.6B-parameter open-weights image model. It supports native 4K generation, pixel-precise editing and up to ten reference images. It has 38% fewer parameters than its predecessor and runs on GPUs with 8 GB of VRAM.
- [NVIDIA releases Nemotron 3.5 Super VL, an open 120B multimodal model](https://www.parallelquant.com/posts/nvidia-releases-nemotron-3-5-super-vl-an-open-120b-multimodal-model-137b1a) (2026-10-15, MarkTechPost): NVIDIA released Nemotron 3.5 Super VL, an open mixture-of-experts model with 120B parameters, 12B active, and a 1M-token context. Its NVFP4 (4-bit) weights run on a single DGX Spark. Two smaller Bolt edge models were released alongside it.
- [Meta releases Muse Spark Open, an 816B-parameter open-weight model](https://www.parallelquant.com/posts/meta-releases-muse-spark-open-an-816b-parameter-open-weight-model-4e8809) (2026-10-15, The Decoder): Meta released Muse Spark Open, an open-weight model with 816 billion parameters that handles text, images, video and audio with up to one million tokens of context. Benchmarks put it roughly on par with the top Chinese open-weight models. Meta and Apple optimized it to run locally on two M5 Ultra Mac Studios using the MLX framework.
- [JetBrains releases Mellum2.1, open 12B coding-agent model](https://www.parallelquant.com/posts/jetbrains-releases-mellum2-1-open-12b-coding-agent-model-97ed5d) (2026-10-08, MarkTechPost): JetBrains released Mellum2.1, an Apache 2.0 mixture-of-experts model with 12B total and 2.5B active parameters. Reinforcement learning in real repositories lifted its SWE-bench Verified score from 2.0 to 47.0.
- [vLLM Semantic Router releases five open Decision 3.0 models](https://www.parallelquant.com/posts/vllm-semantic-router-releases-five-open-decision-3-0-models-4250cb) (2026-10-11, MarkTechPost): vLLM Semantic Router released Decision 3.0, five Apache-2.0 multimodal models ranging from 0.8B to 27B parameters. They answer choice, yes/no and scoring questions about text and images in one call, without generating text.
- [Cloudflare releases open Clef-omni multimodal decision model](https://www.parallelquant.com/posts/cloudflare-releases-open-clef-omni-multimodal-decision-model-423a86) (2026-10-11, MarkTechPost): Cloudflare released Clef-omni, an Apache-2.0 model built on Qwen3-Omni-30B-A3B. It scores text, images, audio and video in a single call, priced at $0.15 per million input tokens. A smaller Clef-flash variant costs $0.038.
- [PoeLLM cryptomining malware infects 3,400+ servers, mostly exposed AI tools](https://www.parallelquant.com/posts/poellm-cryptomining-malware-infects-3-400-servers-mostly-exposed-ai-tool-4b3645) (2026-10-11, Tom's Hardware): Lumen's Black Lotus Labs reports malware called PoeLLM has compromised more than 3,400 servers. Most of the victims were exposed AI tooling such as LiteLLM and Ollama, and the malware is used for cryptomining.
- [Anthropic launches free AI security scans for open-source projects](https://www.parallelquant.com/posts/anthropic-launches-free-ai-security-scans-for-open-source-projects-2d606a) (2026-10-08, The Verge): Anthropic's new OSS Scanner gives opted-in open-source projects periodic vulnerability scans by its strongest models at no cost. Reports are fully model-generated, with no human review or triage, so some may be incorrect or invalid.
- [Liquid AI releases open-weight d1 models that output decisions, not text](https://www.parallelquant.com/posts/liquid-ai-releases-open-weight-d1-models-that-output-decisions-not-text-688e6c) (2026-10-07, MarkTechPost): Liquid AI released d1-3B (text and images) and d1-omni-600M (text with image or audio). Neither model generates text; each returns calibrated, typed answers in one forward pass with zero output tokens. They target real-time decisions.
- [Google releases EmbeddingGemma 2, a 740M open multimodal embedding model](https://www.parallelquant.com/posts/google-releases-embeddinggemma-2-a-740m-open-multimodal-embedding-model-b38b75) (2026-10-06, The Decoder): EmbeddingGemma 2 has 740 million parameters and converts text, images, video, audio and code into vectors. It needs about 191 MB of RAM, runs on-device, and Google says it outperforms some models twice its size. MarkTechPost reports it is built on Gemma 4 and ships under Apache 2.0.
- [Reflection AI unveils Beam, a 501B open-weight coding model](https://www.parallelquant.com/posts/reflection-ai-unveils-beam-a-501b-open-weight-coding-model-1ca568) (2026-10-05, MarkTechPost): Reflection AI's first open-weight model is a 501B-parameter sparse Mixture-of-Experts (MoE) model with 23B active parameters, aimed at coding and agentic work. The company says it matches GLM-5.2 on reasoning with 3 to 4x less inference compute. Apache 2.0 weights are due later in October 2026.
- [Black Forest Labs launches Flux 3 Image with non-destructive editing](https://www.parallelquant.com/posts/black-forest-labs-launches-flux-3-image-with-non-destructive-editing-34a643) (2026-10-02, The Decoder): Flux 3 Image supports multi-step edits that leave the rest of the picture unchanged. Users can compose scenes with bounding boxes and up to ten reference images, with output up to 4K. Open weights are promised in the next few weeks.
- [DeepSeek ships desktop apps for its open-source agent harness](https://www.parallelquant.com/posts/deepseek-ships-desktop-apps-for-its-open-source-agent-harness-d84038) (2026-10-04, MarkTechPost): DeepSeek released official macOS and Windows apps for DeepSeek Harness v0.2, an MIT-licensed agent harness, in preview. It adds a plugin manager, file and code-change review, and scheduled Automation Tasks. It also supports non-DeepSeek models via OpenAI-compatible endpoints.
- [Aleph Alpha releases Kolibri, open-weight 78B English-German MoE model](https://www.parallelquant.com/posts/aleph-alpha-releases-kolibri-open-weight-78b-english-german-moe-model-a4a817) (2026-10-04, MarkTechPost): Kolibri is a 78.1B-parameter Mixture-of-Experts (MoE) model activating only 3.46B parameters per token. It has a 1M-token context, per-request reasoning effort, and Apache 2.0 FP8 weights that run on a single B200 or H200 GPU.
- [NASA and IBM release open-source Lunar Foundation Model](https://www.parallelquant.com/posts/nasa-and-ibm-release-open-source-lunar-foundation-model-c036da) (2026-10-04, The Decoder): NASA and IBM released the Lunar Foundation Model, one of the first open-source AI models for lunar science. It was trained on nearly 2 million tile bundles, mostly from 17 years of Lunar Reconnaissance Orbiter data. It reduces error in predicting polar ice deposits.
- [Physicist used Claude and BootLoops to produce 36 papers in three months](https://www.parallelquant.com/posts/physicist-used-claude-and-bootloops-to-produce-36-papers-in-three-months-0439b8) (2026-10-03, The Decoder): Harvard physicist Matthew Schwartz used the open-source BootLoops harness with Claude to produce 36 manuscripts across 18 fields in three months. The results often became scientifically valuable only after human experts stepped in. Schwartz advises checking everything yourself.
- [Google freezes open-source bug bounty over flood of AI-generated reports](https://www.parallelquant.com/posts/google-freezes-open-source-bug-bounty-over-flood-of-ai-generated-reports-d8cfb4) (2026-10-03, Tom's Hardware): Google suspended product vulnerability submissions to its Open Source Software Vulnerability Reward Program. It cited an influx of invalid, AI-driven reports.
- [Cloudflare releases Clef models for fast structured agent decisions](https://www.parallelquant.com/posts/cloudflare-releases-clef-models-for-fast-structured-agent-decisions-a3c480) (2026-10-02, The Decoder): Cloudflare launched Clef and Clef-flash, built on Qwen and licensed under Apache 2.0, which let AI agents make structured decisions without generating text. Clef-flash returns classifications in about 39 milliseconds, which Cloudflare says is over ten times faster than TypeSafe AI's Jev model.
- [Meta open-sources code for building DIY Muse AI gadgets](https://www.parallelquant.com/posts/meta-open-sources-code-for-building-diy-muse-ai-gadgets-5f24c8) (2026-10-02, The Verge): Meta released SDKs that let hobbyists run its Muse AI agent on devices like ESP32 boards or Raspberry Pi. Suggested projects include an E Ink reminder display, an HDMI stick for big screens, and a small touchscreen device.
- [Alibaba's Qwen releases open-weight image model at 7B params](https://www.parallelquant.com/posts/alibaba-s-qwen-releases-open-weight-image-model-at-7b-params-493c6f) (2026-09-20, The Decoder): Alibaba's Qwen team released Qwen-Image-2.1, an open-weight image generation and editing model that runs on consumer GPUs. It supports transparency layers and up to ten reference images per generation, and the team claims it beats some closed models despite its small 7-billion-parameter size. The research license permits noncommercial use only; commercial use requires a separate Qwen license.
- [Open-source 149M-parameter model tops public sparse embedding benchmark](https://www.parallelquant.com/posts/open-source-149m-parameter-model-tops-public-sparse-embedding-benchmark-242099) (2026-09-19, MarkTechPost): Linkup Research released SPARSEUP, an open-source sparse embedding model built on a 149M-parameter ModernBERT backbone under an Apache 2.0 license. It scores 56.4 nDCG@10 on the BEIR-13 retrieval benchmark, which the company says is the best public result for a sparse encoder under 150M parameters, and reaches over 97% recall in about 380 microseconds per query when paired with the Seismic index.
- [PrismML shrinks a 27B model to 5.9GB with little quality loss](https://www.parallelquant.com/posts/prismml-shrinks-a-27b-model-to-5-9gb-with-little-quality-loss-9f02b2) (2026-09-18, MarkTechPost): PrismML released Ternary Bonsai 2 27B, a ternary-weight compressed version of Qwen3.8 27B that occupies 5.93GB versus 53.80GB for the FP16 original, while retaining 98.2% of the parent model's average performance across 20 benchmarks. It handles text and images with a 262K-token context, released under Apache 2.0.
- [Jina AI releases compact OCR model for low-budget GPUs](https://www.parallelquant.com/posts/jina-ai-releases-compact-ocr-model-for-low-budget-gpus-d86cb4) (2026-09-18, MarkTechPost): Jina AI released jina-ocr-v1, a 3.4B-parameter mixture-of-experts document parser (about 570M active parameters per token) that converts PDFs, scans, tables and charts into Markdown. Built on DeepSeek-OCR with speculative decoding, it scores 91.14 on OmniDocBench v1.6 and 83.4 on olmOCR-Bench, processing 2.57 pages per second on one A100. Weights are on Hugging Face under CC BY-NC 4.0.
- [Alibaba releases Qwen3.8-Omni-Flash, a 1M-context omni-modal model](https://www.parallelquant.com/posts/alibaba-releases-qwen3-8-omni-flash-a-1m-context-omni-modal-model-bba097) (2026-09-18, MarkTechPost): Alibaba's Qwen team released Qwen3.8-Omni-Flash, an omni-modal model with a 1 million token context window that processes audio and video, plans tasks, and calls tools. The company reports it uses about 45.7% fewer tokens on the OmniVideoBench benchmark than prior approaches.
- [Base Labs, Hugging Face, and Goodfire launch open-weight AI safety effort](https://www.parallelquant.com/posts/base-labs-hugging-face-and-goodfire-launch-open-weight-ai-safety-effort-72bf49) (2026-09-17, TechCrunch): Base Labs, a research group spun out of Baseten, launched a partnership with Hugging Face and Goodfire to develop and publish methods for training and monitoring open-weight AI models.
- [Microsoft open-sources TauGrid, a Kubernetes stack for GPU AI workloads](https://www.parallelquant.com/posts/microsoft-open-sources-taugrid-a-kubernetes-stack-for-gpu-ai-workloads-f162e4) (2026-09-17, MarkTechPost): Microsoft's AKS team open-sourced TauGrid, a Kubernetes-native toolkit bundling a CLI, Kueue queueing, KubeRay orchestration, GPU node health monitoring, and observability into one Helm install. It's MIT-licensed and runs on any Kubernetes 1.30+ cluster with GPU nodes.
- [Stanford's Paper2Agent turns research papers into working AI agents](https://www.parallelquant.com/posts/stanford-s-paper2agent-turns-research-papers-into-working-ai-agents-ac9393) (2026-09-16, MarkTechPost): Stanford researchers published Paper2Agent in Nature, a system that converts research papers into validated Model Context Protocol (MCP) tools that can reproduce a paper's results and run on new data. It scored 91.2% on a benchmark of 300 questions spanning 74 papers.
- [China's open-weight AI models now about 4 months behind US frontier](https://www.parallelquant.com/posts/china-s-open-weight-ai-models-now-about-4-months-behind-us-frontier-6e6e02) (2026-09-16, Tom's Hardware): Mozilla's State of Open Source AI report finds China's leading open-weight model, Kimi K3, trails closed US frontier models by roughly four months while costing about 30% as much to use. The models reportedly still lag on some benchmarks.
- [Meta adds MCP server for AI agents to set up WhatsApp Business](https://www.parallelquant.com/posts/meta-adds-mcp-server-for-ai-agents-to-set-up-whatsapp-business-c399fd) (2026-09-15, TechCrunch): Meta released a WhatsApp Business MCP server that lets AI coding agents like Claude, Cursor, Codex, and ChatGPT handle business account setup, messaging templates, testing, and troubleshooting. It targets developers who want an agent to do the configuration work instead of using Meta's dashboards manually.
- [Salesforce launches Koa, a reasoning model for sales tasks](https://www.parallelquant.com/posts/salesforce-launches-koa-a-reasoning-model-for-sales-tasks-043837) (2026-09-15, TechCrunch): Salesforce released Koa, a reasoning-focused AI model built on Nvidia's open-weight Nemotron model. It's trained specifically for sales, marketing, and customer-support tasks rather than general-purpose use.
- [Frontier AI models cost 5x more for a 4-month lead](https://www.parallelquant.com/posts/frontier-ai-models-cost-5x-more-for-a-4-month-lead-d4fefa) (2026-09-15, Ars Technica): A Mozilla-commissioned report reviewed by Ars Technica finds that paying for top frontier AI models delivers roughly a four-month capability lead over free open-weight alternatives, at about five times the cost. The report tracks how quickly open models have been closing the capability gap with proprietary frontier systems.
- [Sakana AI proposes backpropagation alternative for very deep nets](https://www.parallelquant.com/posts/sakana-ai-proposes-backpropagation-alternative-for-very-deep-nets-4bb8ef) (2026-09-14, MarkTechPost): Sakana AI researchers introduced PC-ALM (Augmented Lagrangian Predictive Coding), a layer-local training method that recovers backpropagation-level gradients without a global backward pass. It matches standard backprop across network widths and depths from 8 to 128 layers, and trains 1,000-layer residual networks within about 2 percentage points of backprop accuracy on MNIST. The team released MIT-licensed JAX code.
- [Nvidia open-sources OSMO orchestrator for robot AI training](https://www.parallelquant.com/posts/nvidia-open-sources-osmo-orchestrator-for-robot-ai-training-775161) (2026-09-14, MarkTechPost): Nvidia open-sourced OSMO, the Kubernetes-native workflow orchestrator it uses internally for Project GR00T, Isaac Lab, and Isaac Sim. OSMO lets robotics teams define training, simulation, and hardware-in-the-loop testing in a single YAML file and routes jobs across compute tiers, from GB200 clusters down to Jetson AGX Thor edge devices, under an Apache 2.0 license at version 6.3.1.
- [AllSpark releases Iris open search agents, strongest in their class](https://www.parallelquant.com/posts/allspark-releases-iris-open-search-agents-strongest-in-their-class-ef9331) (2026-09-13, The Decoder): AllSpark released Iris-mini and Iris-pro, open-source search agents built on Qwen models. The team says the models lead benchmarks among open-weight models of similar size, and that training also improved performance on unrelated tasks like general tool use and office work.
- [AWS releases open-source inbox for managing background AI agents](https://www.parallelquant.com/posts/aws-releases-open-source-inbox-for-managing-background-ai-agents-72d8dd) (2026-09-13, MarkTechPost): AWS introduced Pizza Bot, an open-source, self-hosted inbox built on DeepAgents and LangGraph for managing background AI agents. It provides persistent task state, MCP integrations, configurable approval steps, and scheduled workflows across multiple model providers.
- [Cohere releases open-weight translation model for 50 languages](https://www.parallelquant.com/posts/cohere-releases-open-weight-translation-model-for-50-languages-99a07e) (2026-09-11, MarkTechPost): Cohere released North Small Translate, an open-weight mixture-of-experts translation model using 25B of its 218B total parameters per token. It scores 83.6 on the WMT26 benchmark across 50 languages, with weights free for non-commercial use.
- [Anthropic adds automated evaluation tooling for Claude Code plugins](https://www.parallelquant.com/posts/anthropic-adds-automated-evaluation-tooling-for-claude-code-plugins-5c2f96) (2026-09-11, MarkTechPost): Anthropic released a claude plugin eval command that runs a Claude Code plugin against realistic prompts, grades the output using six grader types, and compares results against a run with no plugin loaded. It can be wired into continuous integration (CI) as a gate before shipping.
- [DeepSeek's new V4.1-Flash model cuts memory needs for AI agents](https://www.parallelquant.com/posts/deepseek-s-new-v4-1-flash-model-cuts-memory-needs-for-ai-agents-85f0bc) (2026-09-10, The Decoder): DeepSeek released V4.1-Flash, a 552-billion-parameter multimodal model that activates only 16 billion parameters per token and cuts key-value cache memory to a quarter of its predecessor. On the DeepSWE coding benchmark it narrowly beats Opus 5 and GPT-5.6 Sol, and it ships under the MIT license.
- [Hugging Face launches chat-based tool to run ML experiments](https://www.parallelquant.com/posts/hugging-face-launches-chat-based-tool-to-run-ml-experiments-a3f082) (2026-09-09, The Decoder): Hugging Face launched ML Intern, an AI assistant built into its chatbot that lets users run machine learning experiments through natural conversation without needing ML expertise.
- [Nvidia brings Rust to GPU kernel programming with CUDA Rust](https://www.parallelquant.com/posts/nvidia-brings-rust-to-gpu-kernel-programming-with-cuda-rust-3e447a) (2026-09-08, MarkTechPost): Nvidia announced CUDA Rust, backed by two open-source NVlabs projects: cuda-oxide, which compiles SIMT kernels from Rust MIR through Pliron and LLVM to PTX, and cutile-rs, which JIT-compiles Tile kernels through CUDA Tile IR on stable Rust 1.89+.
- [Browser-based tool collects 50,000+ robot manipulation demonstrations](https://www.parallelquant.com/posts/browser-based-tool-collects-50-000-robot-manipulation-demonstrations-923bc2) (2026-09-07, MarkTechPost): Axis Robotics released AXIS, a browser-based data collection system that moves robot demonstration gathering out of physical labs onto GPU-backed backends. The dataset includes 207 manipulation tasks and 50,129 verified trajectories on a Franka arm. Pretraining on it lifted the pi-0.5 robot policy's LIBERO-Plus score from 83.9 to 88.8, versus 57.5 for a volume-matched control dataset.
- [OpenBMB releases MiniCPM5-2B, an efficient open on-device model](https://www.parallelquant.com/posts/openbmb-releases-minicpm5-2b-an-efficient-open-on-device-model-a39abd) (2026-09-07, MarkTechPost): OpenBMB released MiniCPM5-2B, a 2.52-billion-parameter dense language model with a 131,072-token context window. It averages 53.9 across 34 benchmarks, ahead of Qwen3.5-4B's 51.1, with its strongest gains in tool use, coding agents, and long-context retrieval. The weights, training data, and intermediate checkpoints are released under Apache 2.0, with GGUF builds starting at 1.56GB.
- [Berkeley researchers open-source unified platform for computer-use agents](https://www.parallelquant.com/posts/berkeley-researchers-open-source-unified-platform-for-computer-use-agent-ae8592) (2026-09-06, MarkTechPost): A UC Berkeley-led team released CUA-Lite, an open platform that standardizes the sandboxes, data formats, evaluation, and reinforcement-learning setups used to train and benchmark computer-use agents. It replaces OSWorld's per-task virtual machines with lightweight Docker containers, cutting the per-task footprint from 4.1GB to 0.9GB.
- [Anthropic open-sources a commerce agent blueprint](https://www.parallelquant.com/posts/anthropic-open-sources-a-commerce-agent-blueprint-70cb84) (2026-09-03, MarkTechPost): Anthropic published anthropics/commerce-agents, an Apache-2.0 licensed reference implementation of a shopping agent and a merchant agent. The blueprint bundles scaffolding teams normally rebuild themselves — an agent loop, a tool layer over a product catalog, an approval gate, and an eval suite — aimed at retail, travel, telecom and entertainment use cases.
- [Perplexity open-sources Lily, a fast local inference engine for Apple Silicon](https://www.parallelquant.com/posts/perplexity-open-sources-lily-a-fast-local-inference-engine-for-apple-sil-4d2ec4) (2026-09-03, MarkTechPost): Perplexity open-sourced Lily, a Rust-based inference engine with custom Metal kernels built specifically to run the Qwen3.6-35B-A3B model on Apple Silicon. In testing on a 40-core, 128GB Apple M5 Max chip, it reached roughly 1.23x the prefill throughput and 1.35x the decode throughput of MLX-LM.
- [Nvidia to acquire Hugging Face for $12.9 billion](https://www.parallelquant.com/posts/nvidia-to-acquire-hugging-face-for-12-9-billion-1d3534) (2026-09-03, TechCrunch): Nvidia has agreed to acquire Hugging Face, the platform hosting over 3 million AI models and used by more than 18 million developers, for $12.93 billion. The deal brings one of the most widely used open-source AI model and dataset hosting platforms under the ownership of the world's largest AI chipmaker.
- [Major open-source AI projects replace community PRs with agent teams](https://www.parallelquant.com/posts/major-open-source-ai-projects-replace-community-prs-with-agent-teams-bdaa78) (2026-09-01, Latent Space): Projects including Vercel's AI SDK, Astro, Flue, and tldraw are moving away from accepting drive-by community pull requests, instead using teams of AI agents to triage issues and apply fixes and features. The shift is driven by the sheer volume of contributors these projects now attract.

---
Published by Parallel Quant — https://www.parallelquant.com