Added. The package now identifies itself to OpenRouter with the HTTP-Referer and X-Title headers, which is how the rankings at openrouter.ai/apps group usage. Requests went out anonymous before this, so the traffic counted for nobody. The headers sit on the client, so every request carries them whatever calls it.
Changed. The default model is z-ai/glm-5.3-flash, in place of z-ai/glm-4.6. Input runs $0.071 to $0.388 per million tokens against $0.43 to $0.60, output $0.237 to $1.358 against $1.75 to $2.20, and the context window is 1,310,720 tokens against 204,800. Anyone who set --model or AIHAWK_MODEL sees no change.
The wiki pages that quoted the old model and its prices were rewritten with figures retrieved the same day, including the worked cost comparison.