TokenAssemble

OpenAI · open-weights model

gpt-oss-20b

gpt-oss-20b is a 21B-parameter model (MoE, 3.6B active) by OpenAI. Its smallest quant (MXFP4) needs at least 13.0GB of GPU-addressable memory at 8K context; it loads cleanly at a quality quant on 36 of the 50 machines we track.

Parameters
21B total · 3.6B active
Layers
24
Max context
128K
Attention
HYBRID
License
apache-2.0

Config from openai/gpt-oss-20b

Quants and what they need

Measured file sizes; “needs” = weights + KV cache at 8K context + overhead.

QuantFile sizeNeeds ≥
MXFP412.1 GB13.0 GB

What runs gpt-oss-20b

Every machine we track, graded at its best clean-fit quant.

HardwareBest-quant verdict · llama.cpp · 8K context
Intel Arc A770 16 GBFits · B · MXFP4interactivebeta
Intel Arc B580Tight · C · MXFP4interactive · offloadbeta
AMD Radeon PRO W7900Fits · A+ · MXFP4interactivebeta
NVIDIA GeForce RTX 3060 12 GBTight · C · MXFP4interactive · offloadbeta
NVIDIA GeForce RTX 3070Tight · C · MXFP4interactive · offloadbeta
NVIDIA GeForce RTX 3070 TiTight · C · MXFP4interactive · offloadbeta
NVIDIA GeForce RTX 3080 10 GBTight · C · MXFP4interactive · offloadbeta
NVIDIA GeForce RTX 3080 12 GBTight · C · MXFP4interactive · offloadbeta
NVIDIA GeForce RTX 3080 TiTight · C · MXFP4interactive · offloadbeta
NVIDIA GeForce RTX 3090Fits · A+ · MXFP4interactivebeta
NVIDIA GeForce RTX 3090 TiFits · A+ · MXFP4interactivebeta
NVIDIA GeForce RTX 4060Tight · C · MXFP4interactive · offloadbeta
NVIDIA GeForce RTX 4060 Ti 16 GBFits · B · MXFP4interactivebeta
NVIDIA GeForce RTX 4060 Ti 8 GBTight · C · MXFP4interactive · offloadbeta
NVIDIA GeForce RTX 4070Tight · C · MXFP4interactive · offloadbeta
NVIDIA GeForce RTX 4070 SUPERTight · C · MXFP4interactive · offloadbeta
NVIDIA GeForce RTX 4070 TiTight · C · MXFP4interactive · offloadbeta
NVIDIA GeForce RTX 4070 Ti SUPERFits · B · MXFP4interactivebeta
NVIDIA GeForce RTX 4080Fits · B · MXFP4interactivebeta
NVIDIA GeForce RTX 4080 SUPERFits · B · MXFP4interactivebeta
NVIDIA GeForce RTX 4090Fits · A+ · MXFP4interactivebeta
NVIDIA RTX 5000 Ada GenerationFits · A+ · MXFP4interactivebeta
NVIDIA GeForce RTX 5060 Ti 16 GBFits · B · MXFP4interactivebeta
NVIDIA GeForce RTX 5070Tight · C · MXFP4interactive · offloadbeta
NVIDIA GeForce RTX 5070 TiFits · B · MXFP4interactivebeta
NVIDIA GeForce RTX 5080Fits · B · MXFP4interactivebeta
NVIDIA GeForce RTX 5090Fits · A+ · MXFP4interactivebeta
NVIDIA RTX 6000 Ada GenerationFits · A+ · MXFP4interactivebeta
NVIDIA RTX A6000Fits · A+ · MXFP4interactivebeta
AMD Radeon RX 6800Fits · B · MXFP4interactivebeta
AMD Radeon RX 6900 XTFits · B · MXFP4interactivebeta
AMD Radeon RX 7800 XTFits · B · MXFP4interactivebeta
AMD Radeon RX 7900 XTFits · A+ · MXFP4interactivebeta
AMD Radeon RX 7900 XTXFits · A+ · MXFP4interactivebeta
AMD Radeon RX 9070Fits · B · MXFP4interactivebeta
AMD Radeon RX 9070 XTFits · B · MXFP4interactivebeta
ASUS Ascent GX10Fits · A+ · MXFP4interactivebeta
Beelink GTR9 Pro Ryzen AI Max+ 395 128GBFits · A+ · MXFP4interactivebeta
NVIDIA DGX SparkFits · A+ · MXFP4interactivebeta
Framework Desktop Ryzen AI Max+ 395 128GBFits · A+ · MXFP4interactivebeta
Framework Desktop Ryzen AI Max+ 395 64GBFits · A+ · MXFP4interactivebeta
GMKtec EVO-X2 Ryzen AI Max+ 395 128GBFits · A+ · MXFP4interactivebeta
HP Z2 Mini G1a Ryzen AI Max+ PRO 395 128GBFits · A+ · MXFP4interactivebeta
Mac mini M4 16GBWon't fitbeta
Mac mini M4 24GBFits · B · MXFP4usablebeta
Mac mini M4 Pro 24GBFits · B · MXFP4interactivebeta
Mac mini M4 Pro 48GBFits · A+ · MXFP4interactivebeta
Mac Studio M3 Ultra 96GBFits · A+ · MXFP4interactivebeta
Mac Studio M4 Max 36GBFits · A+ · MXFP4interactivebeta
Mac Studio M4 Max 64GBFits · A+ · MXFP4interactivebeta

Verdicts computed from published specs and measured GGUF sizes — how the math works.