Frontier & Open-Weight Models

First frontier hybrid model combining instant execution with dynamic controllable thinking budget.

Open-weights reasoning powerhouse trained with large-scale GRPO reinforcement learning directly from base model.

Massive scale foundation model emphasizing world knowledge, factual nuance, and creative intuition.

Cost-efficient STEM reasoning model with developer-controllable reasoning effort (Low/Med/High).

Ultra high-speed multimodal reasoning model supporting native video/audio inputs and 1M context.

Highest performance-to-compute ratio open-weight model matching previous 405B capabilities.

Extremely strong multilingual coding and reasoning architecture matching top commercial APIs.

Flagship European enterprise model built for advanced code generation and multilingual reasoning.