Model ID
thinkingmachines/Inkling-NVFP4
Capabilities
Inkling
Inkling is a native multimodal Mixture-of-Experts model designed for reasoning, coding, conversational AI, retrieval-augmented generation, and agentic workflows across text, image, audio, and video. It features 975B total parameters with 41B activated, using a 66-layer architecture with 6,144 hidden size and 64 attention heads. The model activates 6 experts per token across 256 routed experts alongside 2 shared experts, combining hybrid local and global attention with Multi-Token Prediction for efficient inference. Supporting up to a 1M token context window, it is optimized for large-scale multimodal understanding and long-context applications.
About this model
Inkling is a native multimodal Mixture-of-Experts model designed for reasoning, coding, conversational AI, retrieval-augmented generation, and agentic workflows across text, image, audio, and video. It features 975B total parameters with 41B activated, using a 66-layer architecture with 6,144 hidden size and 64 attention heads. The model activates 6 experts per token across 256 routed experts alongside 2 shared experts, combining hybrid local and global attention with Multi-Token Prediction for efficient inference. Supporting up to a 1M token context window, it is optimized for large-scale multimodal understanding and long-context applications.
Model specifications and capabilities are published by the model author and reproduced here from HF Model Card (thinkingmachines/Inkling-NVFP4). Released under Apache 2.0. Further reading: Blog
Deploy Private Inference Endpoint
Choose your hardware and inference engine to get deployment commands and performance benchmarks tailored to your infrastructure
Hardware
B200
1,440 GB
Engine
vLLM
SGLang
Deployment code
Access Restricted
Enter your email to get access to this content
Requires vllm/vllm-openai:inkling or any standard vLLM image v0.26.0+ to support this model configuration.
Performance under load
Tokens/s (total)
Benchmarks measured by Vultr on B200 · vLLM. Throughput and latency vary with concurrency, input length and engine configuration, so treat these as a comparison baseline, not a service guarantee.
NVIDIA HGX B200
Unlock Next-Generation AI Power on Vultr Cloud GPU