Skip to content

[None][doc] Clarify the perf best practice and supported hardware for gptoss #8665

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
merged 4 commits into from
Oct 31, 2025
+5 −5
Merged

[None][doc] Clarify the perf best practice and supported hardware for gptoss #8665

Changes from all commits
Commits
Show all changes
4 commits
Select commit Hold shift + click to select a range
32d9ecd
[None][doc] Clarify the perf best practice and supported hardware for…
dongfengy Oct 25, 2025
f7dadde
Update doc
dongfengy Oct 25, 2025
3d4018e
Update quick-start-recipe-for-gpt-oss-on-trtllm.md
dongfengy Oct 27, 2025
58d6cf1
Update quick-start-recipe-for-gpt-oss-on-trtllm.md
dongfengy Oct 27, 2025
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
10 changes: 5 additions & 5 deletions docs/source/deployment-guide/quick-start-recipe-for-gpt-oss-on-trtllm.md
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -23,12 +23,12 @@ The guide is intended for developers and practitioners seeking high-throughput o

There are multiple MOE backends inside TensorRT LLM. Here are the support matrix of the MOE backends.

| Device | Activation Type | MoE Weights Type | MoE Backend | Use Case |
|------------|------------------|------------------|-------------|----------------|
| B200/GB200 | MXFP8 | MXFP4 | TRTLLM | Low Latency |
| B200/GB200 | MXFP8 | MXFP4 | CUTLASS | Max Throughput |
| Device | Activation Type | MoE Weights Type | MoE Backend | Use Case |
|---------------------- |-----------------|------------------|-------------|--------------------------------|
| B200/GB200/B300/GB300 | MXFP8 | MXFP4 | TRTLLM | Low Latency and Max Throughput |

The default moe backend is `CUTLASS`, so for the combination which is not supported by `CUTLASS`, one must set the `moe_config.backend` explicitly to run the model.
The default moe backend is `CUTLASS`, so for the best possible perf, one must set the `moe_config.backend` explicitly to run the model.
`CUTLASS` was better for max throughput at first but now we have optimized `TRTLLM` moe to be universally faster.

## Deployment Steps

Expand Down