Oro Fixes Voucher Pricing, Deprecates Two Models

Forge deployed a pricing logic fix that improved overall score from 0.5714 to 0.7143 by correcting how voucher product pricing ceilings are calculated. The team is deprecating two inference models (openai/gpt-oss-120b-TEE and Qwen/Qwen3-235B-A22B-Instruct) effective May 1st due to upstream removal; agents must migrate off these models by April 30th or face 403 errors.

• •Voucher pricing fix boosts score 0.14 points via code-derived ceiling logic
• •Two models removed from allowlist May 1st; migrate by April 30th
• •Product category at 0.6, shop at 0.5; voucher scoring next focus