Official site ↗ · origin not confirmed
Serves: Open-weight models through its own API — DeepSeek V4 Pro/Flash, GLM 5.2 / 5.3 Flash, Kimi K2.6 / K2.7 Code, MiniMax M3, Qwen3.8, gpt-oss.
Billing unit: tokens
Converts to tokens: full
Verifiable pre-pay: partial
Pricing / discount: none
Left OpenRouter, so it no longer appears anywhere in the provider table — checked 2026-09-13 against the endpoints of all 445 models on the listing, and it serves none of them. It still sells inference directly, which is why it belongs here: a reader can still buy from it. Two things to know before you do. Inference is billed SEPARATELY from the Enterprise and Pro licences, in arrears, and paid access requires prepayment. The $0.50 per million 'Agent Token Rate' is for the ARIA agent only ('ARIA runs on OpenAI models… This rate applies on top of model API pricing'), not for plain inference (as of 2026-09-18). Free, Pro and Academic plans include some serverless credits; when they run out the account has to switch to pay-as-you-go explicitly. You can cap it: a monthly budget stops requests once reached. 2026-09-18: tabla de modelos alojados con precio: DeepSeek V4-Pro $1,15/$2,55, V4-Flash $0,14/$0,28, GLM 5.2 $0,76/$2,42, GLM 5.3 Flash $0,15/$0,50, Kimi K2.7 Code $0,71/$3,50, Kimi K2.6 $0,65/$3,41, MiniMax M3 $0,23/$0,96, gpt-oss-120b $0,03/$0,17. Lo de 'in arrears' y el prepago no sale en esta pagina: sin confirmar. 2026-09-24: su tabla de modelos alojados pasa de 8 a 30. Nuevos con precio, entre otros: DeepSeek V4-Pro-0813 $1,31/$3,96, V4.1-Flash $0,20/$0,65, Qwen3.8 27B $0,40/$3,00, GLM 5 $1,00/$3,20, Gemma 4 31B $0,10/$0,34, Nemotron 3 Ultra $0,50/$2,15 y JetBrains Mellum2 $0,05/$0,10. Los ref_prices de esta ficha no se mueven. Y el recargo del agente ARIA, leido hoy, es mas amplio de lo que decia esta ficha: 'All plans include an Agent Token Rate of $0.50 per million tokens... for all token types'. Lo de la facturacion a mes vencido y el prepago sigue sin aparecer.