GPT-OSS / GPT-OSS 120B

MXFP4 quantization for GPT-OSS 120B

The published profile is indicative; engineering measurements are completed for the selected model and hardware.

Before → After

Base releaseVerified checkpoint
117.000B

QuantizationMXFP4
Quality retention target

Production artifactArtifact
Lineage and benchmark report

Weights and KV cache

Estimated until measured during engineering

Estimated VRAM

Measured and certified results are delivered with the project.

Runtime packaging

transformers

Not currently orderable