llm-openrouter 0.6

2026-04-20 · Source: Simon Willison's Weblog · Field: Technology & Digital — Artificial Intelligence & Machine Learning, Software Development & Engineering · Depth: Intermediate, quick

Summary

The `llm-openrouter` plugin, version 0.6, for the `llm` command-line utility has been released, introducing a new `llm openrouter refresh` command. This feature allows users to manually update the list of available models from OpenRouter without waiting for the default cache expiration. The update was prompted by the immediate availability of Kimi 2.6 on OpenRouter, enabling quicker access to new models. Kimi 2.6, a new model, was demonstrated generating an interactive HTML and JavaScript UI for an animated pelican, showcasing its capability to produce complex web content with controls for animation speed and wing flap.

Key takeaway

For NLP Engineers integrating new models, the `llm openrouter refresh` command in `llm-openrouter 0.6` is crucial. It ensures you can immediately access and test newly released models like Kimi 2.6 on OpenRouter, bypassing cache delays. This capability is vital for rapid prototyping and staying current with the latest model advancements, allowing you to quickly evaluate new functionalities, such as Kimi 2.6's ability to generate interactive UIs.

Key insights

The `llm-openrouter` plugin update enables immediate access to new OpenRouter models via a refresh command.

Principles

• Prioritize timely access to new models
• Offer manual cache refresh options

Method

The `llm openrouter refresh` command forces an immediate update of the available model list from OpenRouter's API.

In practice

• Use `llm openrouter refresh` for new models
• Explore Kimi 2.6 for UI generation

Topics

• llm-openrouter
• OpenRouter
• Kimi 2.6
• LLM Plugin
• Model Caching

Code references

• simonw/llm-openrouter

Best for: NLP Engineer, AI Engineer, Machine Learning Engineer

Related on AIssential

• Low-Latency Model Router: Automatic LLM Selection Across OpenRouter — Dynamic LLM routing optimizes model selection based on latency, cost, and quality for varied workloads.
• Kimi K2.6 with OpenCode & OpenRouter | Agentic RAG with LangChain, LangGraph & NextJS | 🔴 Live — Kimi K 2.6 demonstrates strong coding capabilities within an agentic environment, but provider choice significantly impacts performance and cost.
• llm 0.32a2 — OpenAI's GPT-5 class models now use `/v1/responses` for interleaved reasoning, enhancing tool call visibility.
• KV Cache Offloading for Context-Intensive Tasks — KV-cache offloading degrades LLM accuracy on context-intensive tasks due to key projection and unreliable landmarks.
• Prompt Caching with the OpenAI API: A Full Hands-On Python Tutorial — Prompt caching reuses repeated LLM input prefixes to enhance API app speed and cost-efficiency.
• How to Run OpenClaw with Open-Source Models — Cost-effective open-source LLMs like Kimi-K2.5 can power OpenClaw effectively, despite minor performance trade-offs.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by Simon Willison's Weblog.