# Open Source — AI updates

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

- [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.
- [Z.ai's GLM-5.3 jumps on coding benchmarks via post-training alone](https://www.parallelquant.com/posts/z-ai-s-glm-5-3-jumps-on-coding-benchmarks-via-post-training-alone-3e6e5b) (2026-08-14, MarkTechPost): Z.ai released GLM-5.3, which reuses the same 743B-parameter GLM-5.2 base model unchanged, with all gains coming from additional post-training. Terminal-Bench 3.0 rose from 4.6 to 28.3 and DeepSWE v1.1 from 46.2 to 66.9, while cybersecurity benchmark CyberGym reached 84.5% and ExploitBench more than doubled to 54.4%; open weights are expected in about two weeks.
- [Google open-sources zero-trust P2P network for AI agents](https://www.parallelquant.com/posts/google-open-sources-zero-trust-p2p-network-for-ai-agents-49bb6c) (2026-08-18, MarkTechPost): Google released Sovereign Agent Mesh (SAM) under Apache-2.0, a zero-config, zero-trust peer-to-peer overlay that lets autonomous agents discover and call each other's MCP tools across cloud, on-prem, laptop, and edge environments without exposing internal endpoints. Identity flows from OIDC into Biscuit capability tokens, enforcing default-deny authorization on every request.
- [Modular's Mojo programming language is now open source](https://www.parallelquant.com/posts/modular-s-mojo-programming-language-is-now-open-source-79b37f) (2026-08-18, Simon Willison): Modular has open-sourced Mojo, its programming language designed as a high-performance alternative to Python for AI and machine learning workloads.
- [Nvidia's TensorRT Model Connect turns checkpoints into C++ inference](https://www.parallelquant.com/posts/nvidia-s-tensorrt-model-connect-turns-checkpoints-into-c-inference-2af09c) (2026-08-18, MarkTechPost): Nvidia released TensorRT Model Connect (TRTMC) in public preview, an Apache-2.0 tool that converts a Hugging Face or local model checkpoint directly into native C++ inference in two commands, with no intermediate ONNX export step. It ships a July 29, 2026 GB300 snapshot covering 105 release profiles across 76 model families.
- [MiniMax releases open-weights AI model for full 5-minute songs](https://www.parallelquant.com/posts/minimax-releases-open-weights-ai-model-for-full-5-minute-songs-d5ea4b) (2026-08-17, MarkTechPost): MiniMax released MiniMax-Music3, an open-weights text-to-music model. Given lyrics with section tags and a structured caption, it generates a complete song up to five minutes long in a single pass, output as 32kHz, 16-bit stereo WAV.
- [DeepSeek releases open-source agent harness with plugin architecture](https://www.parallelquant.com/posts/deepseek-releases-open-source-agent-harness-with-plugin-architecture-e6b072) (2026-08-17, MarkTechPost): DeepSeek AI released DeepSeek Harness v0.1, an MIT-licensed agent harness in developer preview where every capability is implemented as a plugin. It supports four runtime modes, append-only session logs, and provider-agnostic model routing, meaning it isn't locked to DeepSeek's own models.
- [Alibaba releases Qwen 3.8, a 27B open model beating its larger sibling](https://www.parallelquant.com/posts/alibaba-releases-qwen-3-8-a-27b-open-model-beating-its-larger-sibling-23f81d) (2026-08-14, The Decoder): Alibaba's Qwen team released Qwen 3.8, a dense 27-billion-parameter open-weight model under the Apache 2.0 license. It's designed to outperform the larger Qwen 3.7 Plus on coding and office tasks and natively handles up to 262,000 tokens of context, aimed at developers building local and agent-based applications.
- [Zhipu AI releases GLM-5.3, claims top open-weights coding model](https://www.parallelquant.com/posts/zhipu-ai-releases-glm-5-3-claims-top-open-weights-coding-model-166eb7) (2026-08-14, The Decoder): Zhipu AI released GLM-5.3, which it claims is the strongest open-weights coding model, citing a 50% benchmark improvement over its predecessor from post-training alone, according to The Decoder. The company says the model helped security teams find 2,436 vulnerabilities across 269 projects; weights are set to go open source in two weeks.
- [Cactus Compute ships 45M-parameter tool-calling model in 14MB](https://www.parallelquant.com/posts/cactus-compute-ships-45m-parameter-tool-calling-model-in-14mb-340e3d) (2026-08-14, MarkTechPost): Cactus Compute released Needle 2, an open 45-million-parameter model specialized for tool calling, on-device use, and structured data extraction. The full model ships as a single 14MB binary and runs a complete session in about 28MB of RAM, requiring no GPU or NPU. It leads both splits of the Seal-Tools benchmark among models targeting this class of hardware.
- [Writer launches new AI model built on Z.ai's GLM-5.2](https://www.parallelquant.com/posts/writer-launches-new-ai-model-built-on-z-ai-s-glm-5-2-b4ffbf) (2026-08-13, TechCrunch): Writer introduced a new AI model and an upgraded agent harness designed to cut token costs. The model is a post-training variation of Z.ai's open-source GLM-5.2, positioned for deployment-ready enterprise use at a lower price.
- [Ling 3.0 Flash claims top spot among small open models](https://www.parallelquant.com/posts/ling-3-0-flash-claims-top-spot-among-small-open-models-0debd7) (2026-08-13, The Decoder): Ling 3.0 Flash has been released and is being described, per its own benchmark results, as the strongest open-weight model in its size class.
- [DeepSeek ships V4 Pro, open-sources its agent tool, raises API prices](https://www.parallelquant.com/posts/deepseek-ships-v4-pro-open-sources-its-agent-tool-raises-api-prices-299133) (2026-08-13, The Decoder): DeepSeek moved its flagship V4 Pro model out of testing and released its agent software, Harness v0.1, under the MIT license. At the same time it raised API prices, with cache-hit costs jumping sixfold, the largest price increase yet for agent workflows that repeatedly read the same files.
- [Nvidia targets trillion-parameter scale with next Nemotron model](https://www.parallelquant.com/posts/nvidia-targets-trillion-parameter-scale-with-next-nemotron-model-fa5810) (2026-08-12, The Decoder): Nvidia is developing Nemotron 4, a new open-weight model reportedly aiming for roughly one trillion parameters. The scale would put it in the same range as some large models already released by Chinese labs.
- [Liquid AI releases a fast open vision model for edge devices](https://www.parallelquant.com/posts/liquid-ai-releases-a-fast-open-vision-model-for-edge-devices-a831ae) (2026-08-12, Hugging Face): Liquid AI published LFM2.5-VL-3B on Hugging Face, a 3-billion-parameter vision-language model built for faster performance on edge hardware.
- [DeepSeek ships V4 Pro, its largest open-weight model yet](https://www.parallelquant.com/posts/deepseek-ships-v4-pro-its-largest-open-weight-model-yet-5368a8) (2026-08-12, Simon Willison): DeepSeek released the general-availability version of V4 Pro on OpenRouter, a mixture-of-experts model with a 1-million-token context window. The company reports large gains over the preview version on agentic coding and terminal-use benchmarks, and it is now the largest openly available model, surpassing Kimi K2.6 and GLM-5.1.
- [Microsoft's new coding model trails DeepSeek on price and performance](https://www.parallelquant.com/posts/microsoft-s-new-coding-model-trails-deepseek-on-price-and-performance-249194) (2026-08-12, The Decoder): Microsoft released MAI Code 1.1 Flash, a code model for GitHub Copilot said to be 25% more token-efficient and a quarter the cost of its predecessor. In benchmarks it still loses to the cheaper DeepSeek V4 Flash on both price and performance.
- [LTX-2.5 open-weights video model runs on local NVIDIA GPUs](https://www.parallelquant.com/posts/ltx-2-5-open-weights-video-model-runs-on-local-nvidia-gpus-1c81e8) (2026-08-11, MarkTechPost): LTX-2.5 is a new open-weights video generation model built to run on local NVIDIA hardware. It produces 6.8-second video clips with native multishot support and ships with day-one ComfyUI integration.
- [Nvidia's open-weight Nemotron 3.5 Lightning trades size for speed](https://www.parallelquant.com/posts/nvidia-s-open-weight-nemotron-3-5-lightning-trades-size-for-speed-add0a8) (2026-08-11, The Decoder): Nvidia released Nemotron 3.5 Lightning, an open-weight model with 3.6 billion active parameters that reportedly matches OpenAI's gpt-oss-120b on the Intelligence Index despite being roughly four times smaller. It runs at nearly 670 tokens per second, making it the fastest model in its comparison set.
- [Hugging Face and EleutherAI benchmark OCR for AI training data](https://www.parallelquant.com/posts/hugging-face-and-eleutherai-benchmark-ocr-for-ai-training-data-707484) (2026-08-10, The Decoder): The FineBooks project tested 14 open-source OCR models on over 2,000 historical book pages to find the best way to digitize text for AI training. The top model, dots.mocr, hit 97.6% character accuracy at under $2 per thousand pages, though the team says that's not yet accurate enough for scholarly transcription.
- [Meta releases open-weight Muse Glimmer model, Zuckerberg defends open AI](https://www.parallelquant.com/posts/meta-releases-open-weight-muse-glimmer-model-zuckerberg-defends-open-ai-a2666f) (2026-08-10, The Decoder): Meta released Muse Glimmer, a 30-billion-parameter open-weight agentic model from its new Superintelligence Labs that can run on a single consumer GPU once compressed. Alongside the release, Mark Zuckerberg published a roughly 6,500-word essay defending distillation of rival labs' models and arguing for fewer restrictions on US AI labs. An open-weight version of Muse Spark 1.2 is reportedly expected to follow.
- [DeepMind's WeatherNext model forecasts cyclones a day further ahead](https://www.parallelquant.com/posts/deepmind-s-weathernext-model-forecasts-cyclones-a-day-further-ahead-347e01) (2026-08-09, The Decoder): Google DeepMind's WeatherNext model can now predict tropical cyclone tracks and intensity simultaneously, forecasting about a day further ahead than leading operational weather models. DeepMind has open-sourced the code and model weights on GitHub.
- [NVIDIA releases open speech-to-speech AI model with 450ms latency](https://www.parallelquant.com/posts/nvidia-releases-open-speech-to-speech-ai-model-with-450ms-latency-a55b48) (2026-08-09, MarkTechPost): NVIDIA released NemotronLabs VoiceChat 11B, an open-weight full-duplex speech-to-speech model with about 450 milliseconds of turn-taking latency. The model supports live tool calling during voice conversations, letting it invoke external functions mid-dialogue.
- [Open-source tool lets AI agents fork, rewind, and replay runs](https://www.parallelquant.com/posts/open-source-tool-lets-ai-agents-fork-rewind-and-replay-runs-37fa61) (2026-08-08, MarkTechPost): Researchers from Northeastern and Stanford released Shepherd, an MIT-licensed Python runtime that records every agent-environment interaction, including process state and filesystem together, as a Git-like, copy-on-write execution trace. The paper reports 5x faster forking than Docker, over 95% prompt-cache reuse on replay, and a live supervisor that raised CooperBench pair-coding pass rates from 28.8% to 54.7%.
- [Mistral releases open-weight safety classifier matching larger models](https://www.parallelquant.com/posts/mistral-releases-open-weight-safety-classifier-matching-larger-models-e19af4) (2026-08-08, MarkTechPost): Mistral AI released Shieldstral 1.0 3B, an open-weights, policy-adaptive multimodal safety classifier that takes a plain-language policy as input and returns a calibrated safety score without retraining. Built on a 3-billion-parameter base model, it reports 84.9% F1 on text safety, matching a 20-billion-parameter competitor, and runs in 16GB of VRAM under an Apache 2.0 license.
- [NVIDIA open-sources framework that turns an agent into one Python class](https://www.parallelquant.com/posts/nvidia-open-sources-framework-that-turns-an-agent-into-one-python-class-967711) (2026-08-07, MarkTechPost): NVIDIA Labs released NOOA (NVIDIA Object-Oriented Agents), a model-agnostic, open-source Python framework for building AI agents. Instead of separate prompt templates, tool schemas, and workflow graphs, NOOA maps agent state to class fields, available actions to methods, and prompts to docstrings inside a single class.
- [Tencent open-sources team memory hub for AI coding agents](https://www.parallelquant.com/posts/tencent-open-sources-team-memory-hub-for-ai-coding-agents-17a9f4) (2026-08-07, MarkTechPost): Tencent Cloud released TencentDB Agent Memory v2.0, an MIT-licensed, self-hosted memory system that converts conversations, documents, and code into four reusable asset types: Chat Memory, Skill, LLM-Wiki, and Code-Graph. It integrates with Claude Code, OpenClaw, Hermes, and CodeBuddy, and uses access-control lists to decide which agent can see which asset and version.
- [Liquid AI releases small on-device agentic model, open weights](https://www.parallelquant.com/posts/liquid-ai-releases-small-on-device-agentic-model-open-weights-1c6904) (2026-08-07, MarkTechPost): Liquid AI released LFM2.5-2.6B, a 2.69-billion-parameter agentic model that plans, calls tools, and completes multi-step tasks entirely on-device. It handles 131,072 tokens of context, decodes at 220 tokens per second on an Apple M5 Max in under 2.5GB, and ships as open weights in GGUF, MLX, and ONNX formats.
- [Microsoft open-sources unit-test AI agent, beats stock Copilot](https://www.parallelquant.com/posts/microsoft-open-sources-unit-test-ai-agent-beats-stock-copilot-248f6c) (2026-08-07, MarkTechPost): Microsoft open-sourced code-testing-generator, a polyglot unit-test agent shipped in the MIT-licensed dotnet/skills repository. It reads a repo to detect language, test framework, and conventions before writing, running, and validating tests, completing 140 of 152 internal benchmark tasks versus 120 for stock GitHub Copilot on the same underlying model.
- [Amazon, Microsoft, OpenAI, Cursor, Vercel agree on agent plugin standard](https://www.parallelquant.com/posts/amazon-microsoft-openai-cursor-vercel-agree-on-agent-plugin-standard-e8b5d0) (2026-08-07, The Decoder): Amazon, Cursor, Microsoft, OpenAI, and Vercel jointly released Agent Plugins, an open standard defining a single package format for AI agent extensions. Version 1.0.0 uses a plugin.json manifest file and supports both agent skills and MCP servers.
- [Alibaba's Qwen3.8 Max nears Claude Opus 4.8, but Kimi K3 still leads on value](https://www.parallelquant.com/posts/alibaba-s-qwen3-8-max-nears-claude-opus-4-8-but-kimi-k3-still-leads-on-v-8133b3) (2026-08-06, The Decoder): Alibaba's new Qwen3.8 Max scores 56 on the Artificial Analysis Intelligence Index, up 10 points from Qwen3.7 Max's 46, putting it close to Claude Opus 4.8. Moonshot's Kimi K3 still scores higher overall while costing about 25% less.
- [Cloudflare open-sources its internal AI coding agent workspace](https://www.parallelquant.com/posts/cloudflare-open-sources-its-internal-ai-coding-agent-workspace-a9388f) (2026-08-06, Ars Technica): Cloudflare built an internal AI agent workspace to let non-coder employees build software with natural-language prompts, and has now released it as open source. It's positioned as a 'vibe-coding' tool for people without programming backgrounds.
- [DeepMind's WeatherNext AI predicts hurricanes earlier, going open source](https://www.parallelquant.com/posts/deepmind-s-weathernext-ai-predicts-hurricanes-earlier-going-open-source-9d3c8a) (2026-08-06, WIRED): DeepMind's WeatherNext model can predict a hurricane's track and intensity earlier than existing methods using lower-resolution weather data, and will be open-sourced. Researchers reportedly don't yet fully understand the mechanism behind its accuracy.
- [Prime Intellect open-sources an agent harness that beats a human baseline](https://www.parallelquant.com/posts/prime-intellect-open-sources-an-agent-harness-that-beats-a-human-baselin-e98363) (2026-08-06, MarkTechPost): Prime Intellect released Prime Agent, an open-source coding and research harness that treats sub-agents as function calls inside a persistent IPython kernel and lets the agent edit its own prompts, skills, and memory mid-run. Running on Opus 5, it reportedly scored 95.5% Best@1 on ARC-AGI-3, above the cited human expert baseline of 95.4%.
- [Mistral's tiny Shieldstral model matches much larger AI safety filters](https://www.parallelquant.com/posts/mistral-s-tiny-shieldstral-model-matches-much-larger-ai-safety-filters-e81e62) (2026-08-05, The Decoder): Mistral released Shieldstral, a 3-billion-parameter open model that screens AI inputs and outputs for safety violations using natural-language yes/no questions instead of fixed categories. It matches models seven times its size on some benchmarks and can run locally, letting operators set their own safety criteria at runtime.
- [Genspark open-sources a free AI office suite for Mac and Windows](https://www.parallelquant.com/posts/genspark-open-sources-a-free-ai-office-suite-for-mac-and-windows-b0c83e) (2026-08-04, MarkTechPost): Genspark released GenOffice under the Apache 2.0 license, an AI-native office suite covering documents, spreadsheets, slides, and PDFs as five Electron apps sharing one engine. It claims byte-preserving edits, so untouched parts of a file keep their original formatting, and uses an in-house Rust sidecar for spreadsheets. The project is currently in alpha, and AI features require a signed-in Genspark account that consumes credits.
- [Qwen releases new open-weight Max and 27B models](https://www.parallelquant.com/posts/qwen-releases-new-open-weight-max-and-27b-models-4d2792) (2026-08-04, Latent Space): Alibaba's Qwen team released new open-weight models: Qwen 3.8 Max, a 2.4 trillion-parameter model, alongside a smaller 27B variant, aimed at coding and collaborative-work use cases.
- [Y Combinator open-sources its internal multi-agent Slack harness](https://www.parallelquant.com/posts/y-combinator-open-sources-its-internal-multi-agent-slack-harness-288426) (2026-08-04, MarkTechPost): Y Combinator released QM, an MIT-licensed "multiplayer agent harness" it built for internal use across accounting, legal, events, and engineering. Each Slack room gets its own scoped memory, files, permissions, and sandbox, and the same headless core can be driven by Pi, OpenCode, Codex, or Claude Code. It was open-sourced on July 31, 2026.
- [White House shelves Chinese AI ban after Big Tech pushback](https://www.parallelquant.com/posts/white-house-shelves-chinese-ai-ban-after-big-tech-pushback-facf6d) (2026-08-04, The Decoder): The Trump administration discussed sanctions and cloud bans targeting Chinese open-source AI models, according to the New York Times. OpenAI and Anthropic pushed for the restrictions, while Nvidia, Google, and Meta opposed them. Washington has paused the decision for now, with a call expected before Xi Jinping's September visit.
- [New LFM2.5-2.6B model targets local, on-device AI agents](https://www.parallelquant.com/posts/new-lfm2-5-2-6b-model-targets-local-on-device-ai-agents-21e88e) (2026-08-04, Hugging Face): A new model, LFM2.5-2.6B, was published on Hugging Face with 2.6 billion parameters, designed for deploying AI agents locally on-device rather than relying on cloud inference.
- [Cursor open-sources MoE training kernel, 2.37x faster on GB300 racks](https://www.parallelquant.com/posts/cursor-open-sources-moe-training-kernel-2-37x-faster-on-gb300-racks-b4b5c5) (2026-08-04, MarkTechPost): Cursor Research open-sourced Mixture-of-Kittens (MoK), the mixture-of-experts (MoE) training megakernel behind its Composer models. It fuses all MoE communication and computation into a single deterministic kernel and runs up to 2.37x faster than the strongest public baseline, but requires Nvidia Blackwell SM100/SM103 GPUs.
- [Open-weight AI models near frontier capability, but safety lags](https://www.parallelquant.com/posts/open-weight-ai-models-near-frontier-capability-but-safety-lags-088689) (2026-08-04, TechCrunch): A new SaferAI report finds Z.ai's open-weight GLM-5.2 model approaches frontier AI capabilities while lacking key safety mitigations found in closed frontier models. The report renews concerns that powerful open models could outpace governance and safeguards.
- [AMD releases fully open MoE model trained on its own Instinct GPUs](https://www.parallelquant.com/posts/amd-releases-fully-open-moe-model-trained-on-its-own-instinct-gpus-5a1807) (2026-08-01, MarkTechPost): AMD released Instella-MoE-16B-A3B, an open mixture-of-experts language model with 16 billion total parameters but only 2.8 billion active per token, trained from scratch on Instinct MI300X and MI325X GPUs. AMD published the full training pipeline -- weights from every stage, data mixtures, configs, and inference code.
- [Thinking Machines Lab releases open-weight Inkling-Small MoE model](https://www.parallelquant.com/posts/thinking-machines-lab-releases-open-weight-inkling-small-moe-model-65c6ef) (2026-08-02, MarkTechPost): Thinking Machines Lab released Inkling-Small, a 276-billion-parameter mixture-of-experts (MoE) model with only 12 billion active parameters per token, as open weights. It reportedly matches the performance of its larger sibling Inkling at a quarter of the size, and its NVFP4 quantized checkpoint runs on a single Nvidia B300 GPU.
- [MiniMax's open-weight H3 tops an AI video model ranking](https://www.parallelquant.com/posts/minimax-s-open-weight-h3-tops-an-ai-video-model-ranking-e05f5d) (2026-08-03, The Decoder): Chinese AI company MiniMax released weights for its H3 video generation model. It is the first open model to reach the top of an AI video leaderboard, ahead of closed competitors.
- [Alibaba's Qwen3.8-Max launches with 2.4T params, 1M-token context](https://www.parallelquant.com/posts/alibaba-s-qwen3-8-max-launches-with-2-4t-params-1m-token-context-cf4f8f) (2026-08-03, MarkTechPost): Alibaba's Qwen team moved Qwen3.8-Max from preview to general availability, publishing per-token pricing with open weights due next week. The 2.4 trillion-parameter mixture-of-experts model accepts text, image, and video input across a 1 million-token context window; no benchmark table has been published yet.
- [NVIDIA releases Molt, a PyTorch-native agentic RL framework](https://www.parallelquant.com/posts/nvidia-releases-molt-a-pytorch-native-agentic-rl-framework-e6ae80) (2026-08-02, MarkTechPost): NVIDIA open-sourced Molt, a reinforcement learning framework for training AI agents, built with about 8,600 lines of code combining Ray, vLLM, and NeMo AutoModel into a single asynchronous training loop. Agent code stays plain Python with token-exact trajectories, and NVIDIA reports throughput comparable to Megatron-based training stacks.
- [Tencent open-sources AngelSpec framework for faster LLM decoding](https://www.parallelquant.com/posts/tencent-open-sources-angelspec-framework-for-faster-llm-decoding-24c316) (2026-07-30, MarkTechPost): Tencent released AngelSpec, an open-source framework for training speculative-decoding models that speed up large language model (LLM) inference. On its HY3-295B-A21B model, the included drafter delivered a 1.98-2.40x speedup over standard autoregressive decoding.
- [Moonshot AI open-sources MoE parallelism library MoonEP](https://www.parallelquant.com/posts/moonshot-ai-open-sources-moe-parallelism-library-moonep-9c3f0e) (2026-07-30, MarkTechPost): Moonshot AI released MoonEP, an MIT-licensed expert-parallelism (EP) communication library for training large Mixture-of-Experts (MoE) models at scale. It shipped alongside Kimi K3 model weights and technical materials as part of "Kimi K3 Open Day." The library targets more efficient communication between experts during distributed training.
- [Liquid AI open-sources fast, CPU-friendly text encoders](https://www.parallelquant.com/posts/liquid-ai-open-sources-fast-cpu-friendly-text-encoders-da2d12) (2026-07-29, MarkTechPost): Liquid AI released two open-weight bidirectional encoder models, LFM2.5-Encoder-230M and LFM2.5-Encoder-350M, both with 8,192-token context built on the LFM2 hybrid backbone. The 350M model ranks fourth of 14 models on a 17-task GLUE/SuperGLUE/multilingual benchmark suite, and the 230M model completes an 8K-token forward pass on CPU in about 28 seconds.
- [OpenAI open-sources Codex Security CLI to hunt code flaws](https://www.parallelquant.com/posts/openai-open-sources-codex-security-cli-to-hunt-code-flaws-514600) (2026-07-29, The Decoder): OpenAI released Codex Security CLI, an open-source command-line tool that automatically detects and fixes vulnerabilities in code repositories. Previously an internal project called "Aardvark," it has already helped fix more than 3,000 critical security flaws according to OpenAI.
- [30+ firms form Open Secure AI Alliance, skipping OpenAI, Google, Anthropic](https://www.parallelquant.com/posts/30-firms-form-open-secure-ai-alliance-skipping-openai-google-anthropic-f08eb3) (2026-07-27, Tom's Hardware): Over 30 companies, led by Nvidia, have formed the Open Secure AI Alliance to build open-source security models, agent harnesses, and cybersecurity tools. OpenAI, Google, and Anthropic are notably not members. The group argues defenders need locally controlled AI after closed-model safeguards reportedly slowed analysis of the OpenAI-Hugging Face breach.
- [Moonshot AI open-sources Kimi K3 weights and infrastructure](https://www.parallelquant.com/posts/moonshot-ai-open-sources-kimi-k3-weights-and-infrastructure-8a0544) (2026-07-27, The Decoder): Moonshot AI released open weights for its Kimi K3 model along with parts of its training infrastructure. The model nearly matches Western frontier models like Fable 5 and GPT-5.6 Sol on popular benchmarks, but independent testing found notable gaps in cyber and math performance.
- [Open Dreamer reproduces Dreamer 4 world-model pipeline in JAX](https://www.parallelquant.com/posts/open-dreamer-reproduces-dreamer-4-world-model-pipeline-in-jax-7d3014) (2026-07-25, MarkTechPost): A group calling itself Reactor released Open Dreamer, an open implementation of the Dreamer 4 world-model pipeline built in JAX and Flax NNX. Two repositories cover the full stack: a training pipeline with a causal video tokenizer, action-conditioned latent dynamics model, and rollout/FVD scoring, plus a second repo with additional components. The full training recipe was published alongside the code.
- [Microsoft, Meta, Nvidia and 20+ firms back open-weight AI in open letter](https://www.parallelquant.com/posts/microsoft-meta-nvidia-and-20-firms-back-open-weight-ai-in-open-letter-325097) (2026-07-24, The Decoder): More than 20 companies including Microsoft, Meta, and Nvidia signed an open letter advocating for open-weight AI models, as Washington debates restrictions in response to Chinese AI competition. Microsoft is also shifting some of its own products, like Copilot, from external models to its in-house MAI family.
- [Poolside's small open-weight coding model beats larger rivals on benchmarks](https://www.parallelquant.com/posts/poolside-s-small-open-weight-coding-model-beats-larger-rivals-on-benchma-633d64) (2026-07-23, The Decoder): Poolside released Laguna S 2.1, its third coding model in three months, trained to check its own work and revise failed approaches during long agentic sessions rather than relying on raw scale. The compact model reportedly beats several much larger rivals on benchmarks, and Poolside says it solved a math problem open since 1975 for under 10 cents.

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