Find Your Perfect Local AI Model

Compact and fast embedding model. Good balance of quality and speed for semantic search.

Mistral-based vision model. Fast multimodal inference with good image understanding.

Multilingual embedding model supporting 100+ languages. Great for cross-lingual retrieval.

Meta's specialized coding model based on Llama 2. Excellent for code completion and generation.

Google's coding variant of Gemma. Optimized for code completion and generation.

Cohere's model optimized for RAG and tool use. Excellent at following complex instructions.

Cohere's largest model. Optimized for RAG, tool use, and complex multi-step tasks.

Advanced coding model with MoE architecture. Excellent for complex programming tasks.

State-of-the-art reasoning model with chain-of-thought capabilities. Excels at math, coding, and complex reasoning.

Mid-sized distilled R1. Excellent reasoning capabilities for its size.

Compact distilled version of R1. Strong reasoning in an efficient package.

Distilled version of R1 based on Qwen. Great reasoning in a smaller package.

Uncensored Llama 3 fine-tune. General purpose assistant without content restrictions.

Uncensored Mixtral fine-tune. Helpful, harmless, and honest without refusals.

Extended context Llama model. Capable of handling very long documents.

Google's open model built from Gemini research. Available in 2B, 9B, and 27B sizes.

Google's smallest Gemma. Fast and efficient for basic tasks.

Google's efficient 9B model. Great performance in a compact size.

IBM's enterprise-focused coding model. Trained on 116 programming languages.

Meta's powerful 70B model. Excellent reasoning and instruction following with 128K context.

Meta's versatile 8B model with 128K context. Great balance of capability and efficiency.

Efficient smaller models from Meta, perfect for on-device deployment. Available in 1B and 3B sizes.

Ultra-lightweight model from Meta. Perfect for edge devices and fast inference.

Multimodal model that can understand images and text. Available in 11B and 90B variants.

Meta's largest vision model. State-of-the-art multimodal understanding.

Meta's latest and most capable open model. Excellent for general tasks, coding, and reasoning with 128K context.

Visual instruction-tuned model combining CLIP vision with Llama. Great for image understanding.

Large vision-language model. Excellent image understanding and reasoning.

OSS-Instruct trained coding model. Excels at generating clean, documented code.

Highly efficient 7B model that punches above its weight. Great balance of speed and capability.

Mistral's most capable model. Excellent for complex reasoning and enterprise use.

Latest Mistral model optimized for efficiency. Enterprise-grade quality in a compact size.

Mixture of Experts model that activates only 2 experts per token. Fast inference with high quality.

Tiny but capable vision model. Perfect for edge devices and fast image analysis.

State-of-the-art embedding model. Top performance on MTEB benchmarks.

Intel's fine-tuned Mistral for conversational AI. Optimized for dialogue.

High-quality text embedding model. Perfect for RAG, semantic search, and similarity matching.

Nous Research's flagship model. Excellent for roleplay, creative writing, and reasoning.

Powerful Mistral fine-tune trained on large synthetic dataset. Great instruction following.

Compact model with strong reasoning. Good balance between size and capability.

Microsoft's small language model. Surprisingly capable for its size, great for resource-constrained environments.

Microsoft's latest small model with exceptional reasoning. Trained on high-quality synthetic data.

Alibaba's flagship model with excellent multilingual support. Available from 0.5B to 72B.

Efficient 7B version of Alibaba's Qwen 2.5. Great multilingual support and reasoning.

Specialized coding model with excellent code completion and generation. Supports 92 programming languages.

Efficient coding model supporting 92 languages. Great for code completion on consumer hardware.

Hugging Face's tiny but capable model. Perfect for on-device inference and testing.

High-performance embedding model. Top results on MTEB retrieval benchmarks.

Upstage's depth-upscaled model. Strong performance through novel training approach.

Stability AI's coding model. Good for code completion and generation tasks.

Code LLM trained on The Stack v2. Excellent for code completion across 600+ programming languages.

Ultra-compact 1.1B model. Perfect for testing, edge devices, and resource-limited environments.

Uncensored model combining Wizard and Vicuna. Good for creative and unrestricted tasks.

Evol-Instruct trained coding model. Strong on complex programming tasks.

01.AI's bilingual model excelling in English and Chinese. Strong reasoning capabilities.

01.AI's specialized coding model. Strong performance on code generation benchmarks.

Hugging Face's DPO-aligned Mistral. Helpful and aligned assistant.