◆ Section · OSS

Open Source

Open-weight models, frameworks, runtimes — Llama, Mistral, vLLM, llama.cpp, and the rest of the stack.

4 dispatches on file

Smartphone screen showing X-OmniClaw AI agent performing a cross-app task on Android

Open Source18 May 2026 · 4′ read

Oppo Open-Sources X-OmniClaw: An On-Device Android Agent That Runs Camera, Screen, and Voice Locally

Oppo's Multi-X team published X-OmniClaw on GitHub under an Apache 2.0 licence — an edge-native Android AI agent integrating camera, screen recognition, and voice that processes data on-device and calls cloud LLMs only for high-level reasoning, marking the first fully open-source multimodal mobile agent from a major Android OEM.

#oppo#x-omniclaw#android#open-source

Benchmark evaluation chart comparing OpenAI gpt-oss open-weight models against other leading LLMs

Open Source15 May 2026 · 4′ read

gpt-oss: How OpenAI's Apache 2.0 Release Reshaped Open-Weight AI

OpenAI gpt-oss-120b and gpt-oss-20b under Apache 2.0 hit 12 million downloads and set the open-weight benchmark other 2026 model releases compete against.

#openai#gpt-oss#open-weight#apache-2.0

Supply chain attack diagram showing npm package dependency compromise across multiple open source repositories

Open Source15 May 2026 · 3′ read

Mini Shai-Hulud Worm Hits TanStack, Mistral and 170+ npm Packages

Mini Shai-Hulud worm hit TanStack, Mistral AI, and 170+ packages via GitHub Actions cache poisoning, touching 518 million cumulative npm and PyPI downloads.

#supply-chain#security#npm#mistral

Comparison chart of Chinese open-weight coding models GLM-5.1, Kimi K2.6, and MiniMax M2.7 on benchmark tasks

Open Source15 May 2026 · 4′ read

GLM-5.1 vs Kimi K2.6 vs MiniMax: Chinese Coding Models Compared

Benchmarks of GLM-5.1, Kimi K2.6, and MiniMax M2.7 show three distinct winners: GLM for front-end work, Kimi for agentic sessions, MiniMax for cost efficiency.

#glm-5.1#kimi-k2.6#minimax-m2.7#open-weight