Marketplace

Flagship multimodal model — fastest, cheapest GPT-4 for chat, vision and function calling.

Most intelligent model — best for complex reasoning, coding and long-context analysis.

2M token context — the largest context window for document analysis and video understanding.

European open-weight flagship — multilingual, 128K context and top-tier reasoning.

Highest quality text-to-image — precise prompt adherence and photorealistic generation.

Open image generation — self-hostable, fine-tunable and commercially licensed.

Most realistic text-to-speech — clone any voice with 1 minute of audio.

State-of-the-art speech recognition — 99 languages, timestamps and diarization.

Best multilingual embeddings — 128 languages, 512-dim and superior retrieval.

Open reasoning model — o1-level performance at 10x cheaper, fully open weights.