Publisher memory claim
1 × H100 80 GB SXM
80 GB per GPU
Native mixed MXFP4
Compatible MXFP4 runtime + Harmony
Publisher states an 80 GB GPU memory class; workload headroom and full-context concurrency are not guaranteed.
gpt-oss / MoE
gpt-oss-120b
A single large-memory GPU is a useful deployment class
New to model sizes or GPU memory? Start with weights, parameters and quantization →
This model is worth keeping as a text-reasoning baseline when you have one 80 GB-class GPU. The mixed MXFP4 release avoids the multi-node footprint that a naïve BF16 calculation would imply.
117Breported total parameters
5.1Bactive parameters per token
131,072configured token positions
Publisher-reported, rounded model count; not an exact tensor census. Multimodal components and packaging can change the artifact size.
Evaluate it for
Self-hosted text reasoning and controlled studies of scale versus the 20b model.
Choose another path when
Assuming the publisher’s single-GPU memory claim guarantees a full 128K request or high concurrency.
Release context
Why this release matters
It paired 117B total parameters with 5.1B active per token in the same August 2025 release as 20b. The model demonstrates how artifact design can change deployment class without changing parameter count.
Deployment starting points
Which hardware, how many, at what precision?
Start with an exact artifact and an engine that supports it. A configuration below is evidence of a documented path; its device count does not promise a particular throughput or concurrency.
65.25 GB of tensor data in the pinned index
This is the index’s reported tensor payload, in decimal GB. It excludes file headers and runtime memory. Mixed precision, conversion and host/device placement determine how much GPU memory the loaded model needs.
Inspect the exact index metadata →Capacity screen: would the weights fit?
This estimates inference memory for a hypothetical uniform precision. It is useful for rejecting an allocation that is too small. It does not establish a working deployment or estimate training memory.
A uniform-precision GPU count would hide this model’s component layout. Publisher-reported, rounded model count; not an exact tensor census. Multimodal components and packaging can change the artifact size. Use the documented artifact and placement path above; no GPU count is inferred here.
Architecture in practice
What the serving engine has to do
Local and global layers read different spans.
Window length and layer ratio are checkpoint-specific.
Full-context state remains for global layers.
Mechanism schematic based on the pinned model card and configuration. It explains a design principle; it is not a full implementation graph.
Inspect the full checkpoint specifications
| Exact checkpoint | openai/gpt-oss-120b |
|---|---|
| Text model type | gpt_oss |
| Layers | 36 |
| Hidden width | 2,880 |
| Attention / KV heads | 64 / 8 |
| Head dimension | 64 |
| Routed / selected experts | 128 / 4 |
| Configured positions | 131,072 |
| Documented extension | Not recorded for this checkpoint |
| Inputs → output | text → text |
| License metadata | apache-2.0 |
18 sliding attention layers; 18 full attention layers. These counts describe the configured pattern.
Open weights do not imply unrestricted use. Read the applicable license terms linked from the pinned card.
A useful comparison
What else belongs on the shortlist?
20b tests whether the larger model is necessary; Qwen122B adds multimodal input and a different active path.
5 Aug 2025
gpt-oss-20b
A compact reasoning baseline that still earns its place
21B reported total · text
24 Feb 2026
Qwen3.5-122B-A10B
A middle step before the largest Qwen3.5 model
122B reported total · text + image + video
These are editorial comparison candidates. We have not run a matched quality or serving benchmark, so this is not a ranking.
Prepare to explain it
The interview lesson
Explain the gap between 234 GB of hypothetical BF16 weights and a mixed-precision artifact designed for an 80 GB device.
Attention layouts · Quantization · Courses and worked examples
For a deployment evaluation, record exact weights, precision, engine, device count, interconnect, prompt/output lengths and concurrency. Report task success, errors, TTFT, TPOT and useful throughput together.
Sources and citation
Our assessment is editorial judgment based on the linked architecture and deployment evidence. Model facts are publisher-reported or attributed to runtime maintainers; no independent GPU benchmark was run.
AI Infra Interviews, “gpt-oss-120b: release, architecture and deployment”, checked 2026-09-13. Preserve this date and the exact checkpoint when citing.