Model · OpenAI

gpt-oss 120B

117B MoE · Q4_K_M · 58.5 GB · open weights · curated for general, research, agents · repo created 2025-08-04

Too much for a 24 GB card: 61.4 GB at 8K context against 21.1 GB usable. Runs easily on NVIDIA DGX Spark.

What the model needs: weights, context cache and runtime overhead at four reference contexts. The total is the model's own; what differs per machine is the usable memory it has to fit into — VRAM on a graphics card, unified memory on a Mac.

Cache shape from the model's config: 18 of 36 layers cache the full context, each storing K and V for 8 KV heads × 64 dimensions per token at f16; 18 layers see a 128-token sliding window and cache only that. Multiply those four numbers by two bytes and by the context length and you have the cache column.

All memory figures are estimates: measured quantized file size + computed context memory + runtime overhead, with a 12% safety margin on your hardware. Real usage varies with runtime version and settings.

Every tracked GPU and Mac at 8K and 32K context with a full-precision (f16) cache — the same engine as the finder and the GPU pages.

Runs on 1 of 32 tracked devices at 8K context, easily or with ordinary headroom.

GGUF is the quantized single-file format local runtimes load (Ollama, LM Studio, llama.cpp); our memory figures are measured from the linked GGUF file. The original repo holds full-precision safetensors — a much larger download meant for GPUs with far more memory.

More from OpenAI

• gpt-oss 20Bmore popular ▲

Every tracked model →