Realtime AI News
Poolside releases Laguna XS 2.1 open-weight coding model
Poolside has released Laguna XS 2.1, an open-weight coding model for local long-horizon software engineering and agentic workflows, with BF16, FP8, INT4 and NVFP4 checkpoints available on Hugging Face.
Poolside has released Laguna XS 2.1, an open-weight coding model aimed at local long-horizon software engineering and agentic coding workflows.
The model is a 33B-total-parameter MoE system with about 3B active parameters per token, designed for tool use, reasoning, and multi-step coding tasks.
Poolside is providing BF16, FP8, INT4, and NVFP4 checkpoints on Hugging Face. Developers can also use the model through OpenRouter, the Poolside API, Ollama, llama.cpp, vLLM, SGLang, and NVIDIA TensorRT-LLM.
The release also includes DFlash draft models intended to improve local inference speed. That matters for teams exploring coding agents without relying entirely on hosted model APIs.
The broader signal is that coding-agent competition is moving beyond cloud APIs into open weights, local runtimes, and deployment flexibility.
Sources
Why it matters
The release gives developers another open-weight option for private coding agents and local software-engineering workflows.
Nearby Updates
All07/04, 20:57
FactSet Partners with Google Cloud to Integrate Gemini AI into Financial Platform
FactSet, a financial data and analytics provider, announced a partnership with Google Cloud to integrate Gemini AI into its platform. The deal brings generative AI capabilities directly into the workflows of finance professionals.
07/04, 21:18
Relyance AI Warns of AI Agent Privilege Risks, Expands Data-Centric Security Strategy
Relyance AI has highlighted the security risks posed by AI agent privilege management as it expands its data-centric security approach. The warning comes as enterprises increasingly deploy autonomous AI agents that require carefully controlled access to systems and data.
07/04, 21:34
China advances a cybersecurity standard for AI agent deployment
China is moving forward with a cybersecurity standard focused on AI agent deployment, according to a Geopolitechs item surfaced through Google News. The signal matters because enterprise agents increasingly need clear limits around permissions, data access, tool use, and audit trails.
07/04, 21:47
Beijing highlights AI-industry integration through model-data initiatives
At a forum during the 2026 Global Digital Economy Conference, Beijing showcased initiatives aimed at linking AI models, data resources, and industrial applications, including the “model-data resonance” action.