What is Arcee AI
Arcee AI provides hosted inference through an OpenAI-compatible API, including Arcee's Trinity models and selected third-party open models.
Hosted model pricing
As of September 17, 2026, the official rate card lists ten text models. Trinity Large Thinking costs US$0.25 per million input tokens, US$0.80 per million output tokens and US$0.06 per million cached-input tokens. Other listed models include DeepSeek V4 Flash Latest at US$0.14/US$0.28, GLM 5.3 Flash at US$0.15/US$0.50, Inkling Small at US$0.50/US$1.20, DeepSeek V4 Pro 0813 at US$1.32/US$3.96, DeepSeek V4 Pro at US$1.74/US$3.48, GLM 5.2 and GLM 5.3 at US$1.40/US$4.40, Kimi K3 at US$3/US$15 and DeepSeek V4.1 Flash at US$0.22/US$0.66.
The current hosted API supports chat completions and model-specific capabilities such as streaming, tool calls and structured response controls.
Conductor and self-hosting
Arcee Conductor routes requests across supported third-party models without adding a premium to the selected model's API rate. New Conductor accounts are offered US$20 in credits. Trinity weights are also available for self-hosting under their published open licences, avoiding an Arcee inference charge but leaving infrastructure costs to the user.
Official sources: Arcee hosted-model pricing, Trinity model overview, Arcee Conductor pricing.