AI Models
Open-weight models you can run on your own hardware
Code Llama 34B
34BMeta
Meta's specialized coding model built on Llama 2. Excels at code generation, completion, and debugging across multiple p...
Command R
35BCohere
Cohere's RAG-optimized model designed for enterprise search and retrieval tasks. Excels at grounded generation with cita...
DeepSeek R1 14B
14BDeepSeek
Mid-size reasoning model with excellent chain-of-thought capabilities. Balances performance and resource requirements we...
DeepSeek R1 32B
32BDeepSeek
High-performance reasoning model that rivals much larger models. Excellent for complex problem solving, mathematics, and...
DeepSeek R1 70B
70BDeepSeek
The full DeepSeek R1 distilled to 70B parameters. State-of-the-art reasoning that competes with GPT-4 and Claude on math...
DeepSeek R1 7B
7BDeepSeek
Compact reasoning model distilled from DeepSeek-R1, offering strong chain-of-thought reasoning in a small package. Great...
Gemma 2 27B
27BGoogle's largest Gemma model delivering near-Llama-70B quality at half the size. Excellent efficiency-to-performance rat...
Gemma 2 2B
2BGoogle's ultra-lightweight model that runs anywhere. Ideal for on-device AI, edge computing, and rapid prototyping with ...
Gemma 2 9B
9BGoogle's mid-range model offering strong performance across tasks. Uses knowledge distillation techniques for quality th...
Llama 3.2 1B
1BMeta
Ultra-lightweight model perfect for edge devices, mobile phones, and IoT. Surprisingly capable for its tiny size, great ...
Llama 3.2 3B
3BMeta
Small but mighty model that punches above its weight. Runs on almost any hardware and handles conversational AI, summari...
Llama 3.3 70B
70BMeta
Meta's latest instruction-tuned model with exceptional multilingual support and tool use. One of the best open-weight mo...
Mistral 7B
7BMistral AI
The model that proved small models can compete. Mistral 7B uses sliding window attention and grouped-query attention for...
Mixtral 8x7B
46.7B (8x7B MoE)Mistral AI
Mixture-of-experts model that activates only 12.9B parameters per token, giving near-70B quality at far lower compute co...
Phi-4
14BMicrosoft
Microsoft's small language model that achieves remarkable performance through synthetic data training. Excels at reasoni...
Qwen 2.5 14B
14BAlibaba
Mid-range powerhouse from Alibaba with exceptional coding benchmarks. Great balance of capability and resource requireme...
Qwen 2.5 72B
72BAlibaba
Alibaba's flagship open model rivaling GPT-4 class performance. Exceptional at coding, math, and multilingual tasks with...
Qwen 2.5 7B
7BAlibaba
Alibaba's latest 7B model with strong coding and math capabilities. Supports 29+ languages and excels at structured outp...