AI Model Database

Hybrid Mamba-Transformer with 256K context window.

Compact hybrid model with 256K context window.

Alibaba's flagship model with strong multilingual and coding capabilities.

Mid-size Qwen model with excellent coding and reasoning capabilities.

Consumer GPU-friendly model with strong multilingual support.

Efficient multilingual model suitable for edge deployment.

Compact multilingual model for mobile and edge devices.

Ultra-lightweight multilingual model for minimal hardware.

Specialized coding model with state-of-the-art code generation capabilities.

Efficient coding model suitable for consumer hardware.

Lightweight coding model for edge deployment.

Specialized mathematical reasoning model with exceptional problem-solving capabilities.

Vision-language model with exceptional document and image understanding capabilities.

Open code generation model with fill-in-the-middle capabilities.

Efficient code generation model.

Compact code generation model for edge deployment.

Cohere's most capable model with advanced RAG and tool use capabilities.

Efficient model with strong RAG and tool use capabilities.

Compact model with RAG and tool use capabilities, Apache 2.0 licensed.

Databricks' open MoE model with strong coding and reasoning capabilities.

DeepSeek V3 is a 671B parameter MoE model with only 37B activated per token, delivering exceptional efficiency.

Open-source reasoning model matching OpenAI o1 performance on math, code, and logic tasks.

Reasoning capabilities distilled into a 70B model, offering excellent reasoning at manageable size.

Compact reasoning model with impressive performance for its size.

Consumer GPU-friendly reasoning model with strong performance.

Edge-friendly reasoning model suitable for consumer hardware.

Ultra-lightweight reasoning model for minimal hardware.

One of the first truly open large language models.

20B parameter open model from EleutherAI.

Research-focused model designed for interpretability studies.

Google's 27B model with knowledge distillation from larger models.

Efficient 9B model with impressive performance for its size.

Ultra-lightweight model for edge and mobile deployment.

Multimodal model with vision capabilities and extended context.

Efficient multimodal model with vision capabilities.

Compact multimodal model suitable for edge deployment.

Ultra-lightweight multimodal model for minimal hardware.

IBM's enterprise-focused model with strong tool use and safety features.

Compact enterprise model for edge deployment.

Specialized code model with enterprise-grade capabilities.

Efficient code model suitable for enterprise deployment.

Llama 4 Maverick is a 400B parameter Mixture-of-Experts model with native multimodal capabilities, supporting text, image, and video understanding.

Llama 4 Scout features an industry-leading 10M token context window, making it ideal for long-document analysis and code understanding.

Llama 3.3 70B delivers comparable performance to Llama 3.1 405B with significantly reduced computational requirements.

Llama 3.2 90B Vision is Meta's most capable vision model, supporting image understanding and visual reasoning tasks.

Efficient vision-language model optimized for edge deployment while maintaining strong multimodal capabilities.

Lightweight model designed for mobile and edge devices with surprisingly strong performance.

Ultra-lightweight model for the most constrained environments.

Meta's largest open model with state-of-the-art reasoning, tool use, and multilingual capabilities.

Highly capable model balancing performance and efficiency. One of the most popular open models.

Accessible model with strong performance for its size and full multilingual support.

Meta's third generation Llama model with significant improvements in reasoning and coding.

The most popular small open model, excellent for fine-tuning and edge deployment.

Microsoft's Phi 4 with exceptional reasoning and math capabilities from synthetic data training.

Mixture of Experts model with 42B parameters and 128K context.

Compact model with 128K context window for edge deployment.

Mid-size Phi model with strong reasoning capabilities.

Efficient Phi model with strong performance for its size.

Compact model that started the Phi 3 series with impressive capabilities.

Mistral's most capable model with strong multilingual and coding performance.

Latency-optimized model with excellent performance for its size. Apache 2.0 licensed.

Mistral's flagship vision-language model with advanced document understanding.

Efficient vision-language model with strong multimodal capabilities.

Sparse MoE model with 141B total parameters but only 39B active per token.

Groundbreaking MoE model that started the open MoE revolution.

Specialized coding model with fill-in-the-middle capabilities.

12B model with 128K context and strong multilingual capabilities.

The model that proved small models can be highly capable.

NVIDIA's largest open model designed for synthetic data generation and reward modeling.

Efficient model for synthetic data generation.

Enterprise-focused MoE model optimized for SQL and coding tasks.

Stability AI's language model with multilingual support.

Lightweight coding model with fill-in-the-middle support.