Baseten decision models on /v1/decisions

• MiniMax H3 Highspeed, MiniMax H3 Lite - Video output runs at 480P and 768P
• Customer controls - Key-pinned projects let a customer tighten, but not loosen, zero data retention, vendor, and region settings
• Model listings - SDK-compatible listings show only the models an organization can call
• Public models API - Adds each vendor's region
• Flex service tier on Particle: Requests to Particle with service_tier set to flex bill at half of every rate, including cached input and the peak and cache schedule rates. Read more in the Gateway service tiers docs.
• Self-hosted models - An organization's own endpoint is tried before vendor restrictions apply
• Tool schemas - Schemas rooted in oneOf or anyOf work with DeepSeek, Novita, Luminal, and Telnyx
• Upgraded provider request parameter handling.

‍