According to @openclaw on X, OpenClaw’s 2026.3.11 release ships Hunter and Healer Alpha with free 1M-token context models available through OpenRouter, enabling ultra-long-context retrieval and RAG use cases at zero cost for developers (as reported by OpenClaw’s release notes on GitHub). According to OpenClaw, the update integrates Gemini Embedding 2 for improved long-term memory indexing, boosting vector search quality and memory recall in production pipelines. As reported by OpenClaw, GPT 5.4 behavior was tuned to prevent mid-thought stopping, reducing truncation issues in agent loops and code-gen tasks. According to the project’s GitHub release, OpenCode adds Go language support, expanding automated code assistance and test generation beyond Python and JS, while a dedicated security hardening sprint addresses dependency pinning, secret scanning, and sandbox tightening for safer model tooling. For businesses, according to OpenClaw, these changes lower LLM context costs, improve retrieval accuracy, and accelerate multi-language developer workflows, creating opportunities to build durable-memory agents and long-document analytics on top of OpenRouter and Gemini Embedding 2.