AI Models Database

Latest Grok model with 2M context window and fastest output speed of any tracked model (265 tok/s). Available in reasoning and non-reasoning variants.

Latest GPT flagship model with 1M context window. Extended thinking, multimodal input, and document editing capabilities. Outperforms GPT-5.3 Codex on all benchmarks.

Latest Gemini Pro model with thinking capabilities, function calling, code execution, and search grounding. Supersedes Gemini 3 Pro.

Flagship Qwen model with 397B total parameters (17B active via MoE). Hybrid architecture with Gated Delta Networks and 512 experts. Supports 201 languages and thinking/non-thinking modes.

Most intelligent Claude model. Extended thinking with adaptive reasoning. Excels at complex analysis, nuanced content generation, advanced coding, and agentic tasks.

DeepSeek Sparse Attention architecture

Compact reasoning model for efficient multi-step problem solving.

Speed-optimized variant of Imagen 4 for rapid image generation and high-volume tasks. Priced at $0.02 per image.

Frontier-class multimodal model. Performs at or above 90% of Claude Sonnet 3.7. Vision capable with strong reasoning.

Cutting-edge coding model with fill-in-the-middle (FIM) capability. Optimized for code generation, completion, and refactoring.

Open-weight version of FLUX Kontext for research and development. Available on Hugging Face for local deployment.

17B active params, 400B total, multimodal

Midjourney's latest model with new architecture. Features improved text prompt understanding, richer textures, better anatomy rendering, and personalization by default. Includes Draft Mode (10x faster) and Omni Reference.

Most powerful Ideogram model with major leap in visual quality, realism, and creative control. Known for exceptional text rendering in images.

Efficient coding model for development workflows. Cost-effective option for code generation and assistance.

Mid-sized 14B parameter model with text and vision. Strong performance for diverse tasks while remaining efficient.

Frontier multimodal vision model. Excellent for complex image understanding, document analysis, and visual reasoning.

Top-ranked model on Hugging Face Text-to-Image Leaderboard (ELO 1172). Only model capable of generating images with long texts. Supports both raster and vector image generation.

2.5 billion parameter model with improved MMDiT-X architecture. Designed for consumer hardware, requiring only 9.9 GB VRAM.

Cohere flagship for complex enterprise tasks

OpenAI's text-to-image model integrated natively into ChatGPT. Significantly improved prompt understanding and text rendering compared to DALL-E 2.

Intelligent routing to best model for task

High-quality voice synthesis

State-of-the-art embeddings for search

Mixtral on Groq LPU hardware

Chain-of-thought search model

Amazon proprietary model for AWS users

Best-in-class transcription model