AI Models

Luma’s first unified understanding and generation model, a decoder-only autoregressive transformer that interleaves text and images for multimodal reasoning, visual editing, and image generation

Next-generation foundational image generation model unifying text-to-image generation and image editing with professional typography rendering, native 2K resolution support, and 1k-token prompt instructions.

Most capable and efficient frontier model for professional work, with native computer-use capabilities, tool search, and extreme reasoning

An efficient Mixture-of-Experts language model with 671B total parameters, featuring DeepSeek Sparse Attention (DSA) for enhanced reasoning, agentic performance, and long-context efficiency, comparable to frontier models like GPT-5.

State-of-the-art image generation and editing model (Gemini 3.1 Flash Image) combining Pro-level quality, advanced world knowledge, real-time web search grounding, subject consistency, and precise text rendering with Flash-level speed.

Grok 4.20 Beta: xAI's frontier multimodal model with native 4-agent multi-agent collaboration system (Grok, Harper, Benjamin, Lucas) for real-time debate, fact-checking, and reduced hallucinations

A 1-trillion parameter Mixture-of-Experts (MoE) multimodal model featuring native vision capabilities, complex reasoning, and Agent Swarm support for parallel sub-agent coordination.

Gemini 3.1 Pro is the next iteration in the Gemini 3 series of models, a suite of highly capable, natively multimodal reasoning models.

Lightweight 9B-parameter open-source multimodal vision-language model optimized for local deployment, low-latency inference, and edge/consumer hardware; part of the GLM-4.6V series with native multimodal function calling, strong visual understanding, and long-context capabilities

Paid, enhanced version of GLM-4.6V-Flash multimodal model with higher capacity and stability; supports native multimodal tool calling, vision-language tasks, and long-context processing

Open-source multimodal vision-language model with native function calling, state-of-the-art visual understanding and reasoning at its scale, long-context multimodal processing, and support for interleaved image-text generation and agentic workflows

Specialized coding variant of the GLM-5 flagship model, optimized for advanced programming, complex code generation, agentic development workflows, and superior performance in software engineering tasks

Flagship Mixture-of-Experts foundation model designed for Agentic Engineering, complex systems reasoning, long-horizon agent tasks, advanced coding, and low hallucination rates

A proprietary Optical Character Recognition (OCR) model specialized in complex tables, forms, handwritten content, and multi-page PDFs, outputting high-fidelity Markdown or HTML structures.

Frontier-class multimodal model optimized for enterprise-grade performance, high-speed reasoning, and long-context document understanding.

Open-weight, general-purpose, flagship multimodal and multilingual model.

Released in Feb 2026, Opus 4.6 is Anthropic's frontier model featuring 'Adaptive Thinking' and 'Agent Teams' capabilities. It offers a 1M token context window (in beta) and is optimized for complex reasoning, deep research, and autonomous coding tasks.

Anthropic's intelligent flagship model optimized for complex agents, software engineering, and computer use. It features 'extended thinking' for deep reasoning and achieves state-of-the-art performance on coding benchmarks like SWE-bench Verified.

GPT-5.2 is a more reliable and capable AI model with stronger reasoning, longer context understanding, and improved real-world usability.