Know your tools
A maintained field index of leading language models, their capabilities, and the providers behind them.
by Moonshot AI
Kimi K3 is Moonshot AI’s natively multimodal mixture-of-experts model for long-horizon coding, knowledge work, and reasoning. Moonshot reports 2.8 trillion total parameters, 16 active experts out of 896, and a one-million-token context window.
by Runway
Gen-4.5 is Runway’s video-generation model for text-to-video and image-to-video workflows. Runway’s API model directory lists the identifier gen4.5 with text or image input and video output.
by Meta
Llama 4 Maverick is Meta’s natively multimodal, mixture-of-experts model with 17 billion active parameters, 400 billion total parameters, and 128 experts. It accepts multilingual text and images and produces multilingual text and code.
by OpenAI
Maintained facts and independent coverage are being assembled from primary sources.
by OpenAI
Maintained facts and independent coverage are being assembled from primary sources.
by Anthropic
Maintained facts and independent coverage are being assembled from primary sources.
by Anthropic
Maintained facts and independent coverage are being assembled from primary sources.
by Google
Maintained facts and independent coverage are being assembled from primary sources.
by Meta
Maintained facts and independent coverage are being assembled from primary sources.
by Mistral AI
Maintained facts and independent coverage are being assembled from primary sources.
by Mistral AI
Maintained facts and independent coverage are being assembled from primary sources.
by Stability AI
Stable Image Ultra is Stability AI’s hosted text-to-image service built from its advanced image models, including Stable Diffusion 3.5. The official API positions Ultra for high-detail, photorealistic output, typography, and complex compositions.
by Black Forest Labs
FLUX.2 Max is Black Forest Labs’ highest-quality hosted FLUX.2 image configuration. Its official documentation lists multi-reference generation and editing, typography and color control, and optional grounding search for current information.
by Luma AI
Ray 2 is a Luma video-generation model exposed by the Dream Machine API. The generation endpoint accepts a prompt, optional keyframes and looping, and supports multiple aspect ratios, resolutions, and five- or nine-second durations.
by Adobe
Adobe Firefly's latest image model, with native 4 MP resolution and support for image generation and natural-language Instruct Edit.
by Adobe
Adobe's newer Precise Composite model, described as outperforming the previous MD model in object-edge accuracy, lighting realism, and contextual scene generation.
by Allen Institute for AI
Base model with strong results in programming, reading comprehension, and math problem solving, extended-context performance, and suitability for reinforcement learning setups.
by Allen Institute for AI
Smaller, lightweight base model designed to run on a wider range of hardware while delivering competitive performance.