Summary

Update the Venice AI provider with the latest models and improve discovery reliability.

Changes needed:
Add Kimi K2.5 to the static catalog: currently missing, so users don't see it when API discovery fails
Update default model: Kimi K2.5 offers reasoning + vision and is fully private, making it a better default
Add new models: claude-opus-4-6, zai-org-glm-4.7-flash, zai-org-glm-5, minimax-m25, and claude-sonnet-4-6
Update privacy modes: align kimi-k2-thinking and minimax-m21 with Venice API values
Improve discovery: increase timeout for reliability, filter offline models
Update docs: refresh model tables and recommendations

Steps to reproduce

1. Run openclaw onboard --auth-choice venice-api-key
2. Notice Kimi K2.5, GLM 5, MiniMax M2.5, and Claude Sonnet 4.6 are missing from model selection
3. Or: disconnect from internet and run openclaw models list | grep venice — static catalog is missing recent models

Expected behavior

• All current Venice models should appear in the model picker
• Static catalog should include latest models as fallback when API is unreachable
• Default model should be a capable option with reasoning + vision

Actual behavior

• Kimi K2.5, GLM 5, MiniMax M2.5, and Claude Sonnet 4.6 are missing from model selection
• Static catalog is missing recent models when API is unreachable
• Default model is llama-3.3-70b which lacks reasoning and vision capabilities

OpenClaw version

2026.2.9

Operating system

All (macOS, Linux, Windows)

Install method

No response

Logs, screenshots, and evidence

Impact and severity

No response

Additional information

No response