Model Anatomy See what's inside open-weight AI models

Can I run Swift-1.5-Flash-Next-Heretic-ReverseEngineer on an NVIDIA H200?

Yes, just

Its IQ4_NL file (109 GB) fits in the 127 GB an NVIDIA H200 can use, with room for ≈ 256K tokens of context.

Full analysis of Swift-1.5-Flash-Next-Heretic-ReverseEngineer Everything that runs on NVIDIA H200 Get the IQ4_NL file on Hugging Face
llama-server -hf trubisky/Swift-1.5-Flash-Next-Heretic-ReverseEngineer:IQ4_NL

Files and context on NVIDIA H200

Each file of Swift-1.5-Flash-Next-Heretic-ReverseEngineer (and of GGUF versions the site has analyzed), and how much context fits next to it, with the context cache in 16-bit or 8-bit.

FileSizeOn NVIDIA H200 Context (16-bit)Context (8-bit)
IQ4_NL109 GBTight256K256K
IQ4_XS94.7 GBFits256K256K
IQ3_S84.6 GBFits256K256K

Swift-1.5-Flash-Next-Heretic-ReverseEngineer on other hardware

Popular models that run on NVIDIA H200

Assumes 90% of the card's memory is usable and at least 4K tokens of context for one conversation, with the context cache in 16-bit unless noted. Fits leaves at least 20% of that memory for context; Tight leaves less.