Model library

Code-specialized Qwen2.5 with state-of-the-art code generation, reasoning, and fixing across many languages.

Meta's compact Llama 3.2 models (1B and 3B) — efficient, on-device-friendly assistants.

Qwen2.5 is Alibaba's latest series of multilingual LLMs with strong instruction following, long context, and tool use.

Meta's Llama 3.1 — a strong, widely-used open model family (8B / 70B / 405B) with a 128K context.

Gemma 2 — Google's open models (2B / 9B / 27B) built from the same research as Gemini.

Mistral 7B — a fast, efficient, fully open (Apache-2.0) model that punches well above its size.

Mixtral 8x7B — a sparse mixture-of-experts model offering strong quality at high inference efficiency.

A tiny, fast vision-language model that runs almost anywhere — captioning and visual Q&A on the edge.

A state-of-the-art large embedding model for English retrieval and semantic search.

StarCoder2 — open code models trained on permissively-licensed source across 600+ languages.

A high-performing open text embedding model with a large context window, fully open (Apache-2.0).

DeepSeek-R1 — an open reasoning model trained with reinforcement learning, with strong math and coding.

Phi-4 — Microsoft's 14B model with outsized reasoning ability for its size, MIT-licensed.

Whisper — a robust multilingual speech-to-text model for transcription and translation.

Stable Diffusion 3.5 Large — a powerful 8B text-to-image model with improved typography and photorealism.

FLUX.1 [schnell] — a fast, fully open (Apache-2.0) text-to-image model that generates high-quality images in just 1–4 steps.

FLUX.1 [dev] — a 12B text-to-image model with exceptional quality and prompt adherence (non-commercial license).

Stable Diffusion XL — the widely-used open text-to-image model with a huge ecosystem of fine-tunes and LoRAs.