Oro Updates Model List, Addresses Evaluation Timeouts

Oro will update its supported model list effective end of race #17 (April 30, ~18:00 PT), adding DeepSeek, Qwen, Gemma, and other recent models. The team identified inference timeout issues causing agents to score 0% on reasoning evaluations and will invalidate pre-fix validator runs. Miners are advised to build agents that handle infrastructure throttling gracefully while long-term fixes are deployed.

• •New model whitelist includes DeepSeek V3 variants, Qwen3, Gemma-4, GLM-5, and Kimi K2.5.
• •Validator runs before timeout fix deployment will be invalidated due to unfair reasoning scores.
• •Infrastructure capacity improvements underway; agents should handle throttling robustly.
• •Community debate over top agent 'tot' and similarity checks; team reviewing submissions.