Reverse Engineering
Select the model you want to run. We'll tell you what hardware you need.
Llama 4 Behemoth
TextFlagship 2T foundation model, 16 experts
Llama 4 Maverick
TextHigh-efficiency MoE, 128 experts, 1M context
Llama 4 Scout
TextConsumer flagship MoE, 16 experts, 10M context
Mistral Large 3
TextGranular MoE flagship, 256K context
Mistral Large 3 NVFP4
TextFP4 quantized version for NVIDIA NIM
Ministral 3 14B
TextDense edge flagship with vision
Ministral 3 8B
TextBalanced edge model with vision
Ministral 3 3B
TextLightweight mobile model with vision
Grok-3 Mini
TextEfficient reasoning model with real-time tools
Llama 3.3 70B
TextRefined Llama 3 with superior following
Llama 3.2 90B Vision
TextMultimodal with image understanding
Llama 3.2 11B Vision
TextCompact multimodal model
Llama 3.2 3B
TextMobile-optimized small model
Llama 3.2 1B
TextUltra-light edge deployment
Llama 3.1 405B
TextFrontier-class open model. Requires datacenter hardware.
Llama 3.1 70B
TextEnterprise-grade intelligence
Llama 3.1 8B
TextBest small model for most tasks
Qwen 2.5 72B
TextTop-tier reasoning and coding
Qwen 2.5 32B
TextThe "Goldilocks" model - great balance
Qwen 2.5 14B
TextStrong mid-size model
Qwen 2.5 7B
TextEfficient general purpose
Qwen 2.5 3B
TextLightweight and fast
Qwen 2.5 1.5B
TextEdge deployment ready
Qwen 2.5 0.5B
TextSmallest Qwen variant
Qwen 2.5 Coder 32B
TextState-of-the-art code generation
Qwen 2.5 Coder 14B
TextStrong coding in smaller package
Qwen 2.5 Coder 7B
TextEfficient code assistant
Qwen 2.5 Coder 3B
TextLightweight coder
Qwen 3 Max (Thinking)
TextFlagship reasoning model with "System 2" thinking mode
Qwen 3 235B (MoE)
TextOpen weights flagship, highly efficient experts