Home · 4

Claude Code v2.1.302 adds model control for auto-mode classifiers, caps image reads at 128MB, stabilizes plugin workflows, and patches session persistence bugs.

Meta has released Muse Spark Open, an 816B parameter open-weight MoE model designed for agentic coding and multimodal tasks with up to 1M context.

v2.1.301 adds configurable retry timeouts, OpenTelemetry tracking, vim bracket navigation, and critical fixes for agent hooks and terminal stability.

Reflection releases Beam, a 501B open-source MoE LLM featuring 1M context, 23B active parameters per token, and Apache 2.0 licensing.

NVIDIA and AI4Bharat have released a 2.8 trillion token Indic language dataset and an open GPU-accelerated pipeline for synthetic data generation.

n8n integrates Google’s Gemini 4 Argon across its Model Context Protocol server, BYOK Assistant, and standalone AI agents.

Ollama v0.40.5 releases NVIDIA Nemotron 3.5 Super and Bolt, adds cloud decision models, fixes OpenAI Responses tool call breaks, and improves cloud error report

Anthropic's latest Claude Code update adds idle compaction, granular mod hooks, safer plugin evaluations, and several stability fixes.

Addresses token accounting regressions, stabilizes headless execution paths, and adds granular subagent configuration options.

Liquid AI's open-weight 24B MoE reasoning model built for local agent execution, featuring 2B active parameters and explicit chain-of-thought generation.