Models and Pricing
Latest Models
Section titled “Latest Models”The models below are currently available through the relaxAI API. All pricing is based on usage per 1M tokens.
Use the model names exactly as shown when specifying the model parameter in your API requests.
Mistral-7b-embedding
Embedding Model
GPT-OSS-120b
Cutting edge reasoning model by OpenAI, with state of the art performance and advanced agentic functionality.
Llama-4-Maverick-17B-128E
*Sunsetting* A multi-modal native model by Meta, that allows for text and image understanding for non-agentic usecases.
Kimi-K25
*Temporary* Multimodal model with strong agentic capabilities designed for complex coding problems
GLM-46
Reasoning model designed to excel at complex coding problems and advanced agentic functionality.
Nemotron-3-Super
A reasoning model from NVIDIA, best for building specialized AI agents and long-context workflows.
DeepSeek-V4-Pro
Frontier MoE model designed for reasoning, coding, and long-context agentic tasks (up to 1M tokens)