Qwen3.5 is Black Magic
General
OpenClaw Developer acqui-hired Peter Steinberger, the Austrian solo developer behind OpenClaw, was acqui-hired by OpenAI for an undisclosed multi-billion-dollar amount. Sam Altman: “The future will be strongly multi-agent.” OpenClaw will remain open source under an independent foundation. Link: https://www.leanware.co/insights/openai-openclaw-acquisition
State-of-the-Art Models on Local Hardware With the AMD Ryzen AI Max+, it is now possible to run models with one trillion parameters entirely locally on a cluster of four desktop PCs. Link: https://www.amd.com/en/developer/resources/technical-articles/2026/how-to-run-a-one-trillion-parameter-llm-locally-an-amd.html
OpenClaw + Local AI = ❤️ The open-source software Ollama, with the new 0.17 update, makes local models up to 40% faster and now sets up OpenClaw with a single command. This makes local AI not only more powerful but also significantly more accessible. Link: https://www.phoronix.com/news/ollama-0.17
New Models
Alibaba Qwen 3.5 Alibaba released the open-source LLM Qwen 3.5 in several sizes. The 35B model outperforms Claude Sonnet 4.5 and GPT-5 Mini on coding benchmarks. The 122B model reaches GPT-5 High levels on GPQA Diamond and HLE. And the 9B model even surpasses OpenAI’s GPT-OSS-120B, a model with 13 times more parameters. All three can be run locally on consumer hardware. Link: https://venturebeat.com/technology/alibabas-small-open-source-qwen3-5-9b-beats-openais-gpt-oss-120b-and-can-run
MiniMax 2.5 MiniMax released the freely available open-weights model M2.5 (229B parameters, MIT license), which outperforms Anthropic’s Claude Opus 4.6 on coding benchmarks at a fraction of the cost. Running at 100 tokens per second, it is twice as fast as other frontier models, making it ideal for local agent workflows. Link: https://www.datacamp.com/blog/mini-max-m2-5
Zhipu AI GLM-5 The Chinese startup Zhipu AI released GLM-5, an open-source model with 744 billion parameters under the MIT license. It achieves 77.8% on SWE-bench Verified (a coding benchmark), outperforming Google Gemini 3 Pro while also having the lowest hallucination rate among all tested models. Link: https://www.reuters.com/technology/chinas-ai-startup-zhipu-releases-new-flagship-model-glm-5-2026-02-11/
Additional Interesting Links
- Alibaba Qwen 3.5: https://the-decoder.de/alibabas-offene-qwen-3-5-modellreihe-liefert-mehr-leistung-bei-weniger-rechenaufwand/ https://huggingface.co/Qwen
- MiniMax 2.5: https://artificialanalysis.ai/models/minimax-m2-5
- Zhipu AI GLM-5: https://huggingface.co/blog/mlabonne/glm-5 https://llm-stats.com/blog/research/glm-5-launch