Models
Deploy and scale models on your GPU infrastructure of choice with NVIDIA NIM inference microservices
models
Mistral Large 4 from Mistral AI, a 1T-total / 52B-active granular MoE for text and image input with reasoning, function calling and structured outputs, served in NVFP4 with DSpark speculative decoding.Free EndpointMistral Large 4
Nemotron 3.5 Super early-access reasoning model for agentic, coding, tool-calling, and multimodal workloadsFree Endpointnemotron-3.5-super-vl-120b-a12b
Preview of StepFun's Step 5: ~600B-total / 27B-active multimodal MoE (text and image in, text out) with 1M context, reasoning and tool calling, served in NVFP4.Free EndpointStep 5 Preview
552B MoE, 8B active params with native multimodal support and lower API cost using smaller KV cacheDownloadableFree Endpointdeepseek-v4.1-flash
Multimodal 320B-total / 18B-active MoE with hybrid KDA and sparse MLA attention, native FP8 weights, reasoning and tool calling.Free Endpointglm-5-3-flash
284B MoE (13B active) model ideal for long-context workloads optimized for coding, chat, and agentic workflowsDeprecatedFree Endpointdeepseek-v4-flash-0731
Fastest 30B A3B MoE model with leading domain accuracy for specialized agentic tasksDownloadableFree Endpointnemotron-3.5-lightning-30b-a3b
Efficient 33B MoE for local, long-horizon agentic coding and terminal tasksFree Endpointlaguna-xs-2.1
Open, efficient hybrid Mamba-Transformer MoE with 1M context, excelling in agentic reasoning, coding, planning, tool calling, and moreDownloadableFree Endpointnemotron-3-ultra-550b-a55b
Open, efficient hybrid Mamba-Transformer MoE with 1M context, excelling in agentic reasoning, coding, planning, tool calling, and moreDownloadableFree Endpointnemotron-3-super-120b-a12b
Smaller Mixture of Experts (MoE) text-only LLM for efficient AI reasoning and mathDownloadableFree Endpointgpt-oss-20b
Items per page
of 1 pages