Source: https://seerepo.com/compare/Dao-AILab/flash-attention/vs/NVIDIA/TensorRT-LLM
Markdown source: https://seerepo.com/compare/Dao-AILab/flash-attention/vs/NVIDIA/TensorRT-LLM.md
Description: flash-attention and TensorRT-LLM share 7 technologies: apache-tvm-ffi, NVIDIA, quack-kernels, torch-c-dlpack-ext, torch, einops, and triton. Both are written in Python.

# flash-attention vs TensorRT-LLM

flash-attention and TensorRT-LLM share 7 technologies: apache-tvm-ffi, NVIDIA, quack-kernels, torch-c-dlpack-ext, torch, einops, and triton. Both are written in Python.

- [Dao-AILab/flash-attention](https://seerepo.com/r/Dao-AILab/flash-attention): Fast and memory-efficient exact attention
- [NVIDIA/TensorRT-LLM](https://seerepo.com/r/NVIDIA/TensorRT-LLM): TensorRT LLM provides users with an easy-to-use Python API to define Large Language Models (LLMs) and supports state-of-the-art optimizations to perform inference efficiently on NVIDIA GPUs. TensorRT LLM also contains components to create Python and C++ runtimes that orchestrate the inference execution in a performant way.

## What they share and what sets them apart

- In common: apache-tvm-ffi, NVIDIA, quack-kernels, torch-c-dlpack-ext, torch, einops, triton
- Only in flash-attention: triton-windows
- Only in TensorRT-LLM: FastAPI, blobfile, h5py, pyzmq, OpenAI, anyio, uvicorn, starlette, Rich, matplotlib, Plotly, accelerate, cuda-python, flash-linear-attention, flashinfer-python, nvidia-cudnn-frontend, openai-codex, openai-harmony, optimum, rouge, rouge-score, sentencepiece, tiktoken, torchao, transformers, diffusers, safetensors, torchvision, datasets, flash-attn, xgrammar, librosa, MCP, mistral-common, partial-json-parser, tritonclient, evaluate, llguidance, lm-eval, peft, jsonlines, pillow, jinja2, mako, Pydantic, msgspec, jsonschema, opentelemetry-api, opentelemetry-semantic-conventions-ai, psutil, prometheus-client, boto3, OpenTelemetry, opentelemetry-sdk, prometheus-fastapi-instrumentator

## Side by side

| | flash-attention | TensorRT-LLM |
|---|---|---|
| Size | 139,521 lines written, 150 code files | 2,884,778 lines written, 3,120 code files |
| Who writes it | 1.7% of its commits signed by AI. With Claude Code, Cursor, and GitHub Copilot | 1.7% of its commits signed by AI. With Claude Code, CodeRabbit, GitHub Copilot, and Cursor |
| Activity | Last push November 16, 2026, 1,613 commits, 25,304 stars | Last push November 8, 2026, 10,997 commits, 15,038 stars |
| Tests | 80 test files. 32% of its code files have a test that imports them | 2,703 test files. 58.6% of its code files have a test that imports them |

- Size: Lines written by people and agents, without generated ones; code files, without tests.
- Who writes it: A commit counts as AI's when it carries its signature: a Co-Authored-By line, its author name or the merge of its branch.
- Activity: From GitHub and its history, on the day SeeRepo read it.
- Tests: The files that test the code, and the share of code files a test imports.

## Frequently asked questions

### Which is bigger, flash-attention or TensorRT-LLM?

TensorRT-LLM, with 2,884,778 lines written; flash-attention has 139,521.

The proof: [NVIDIA/TensorRT-LLM's code at commit b433473](https://github.com/NVIDIA/TensorRT-LLM/tree/b433473ba0d0d49764f4d2fb3d45c82716b946dc)

### Which has more GitHub stars, flash-attention or TensorRT-LLM?

flash-attention, with 25,304 stars; TensorRT-LLM has 15,038.

The proof: [Dao-AILab/flash-attention on GitHub](https://github.com/Dao-AILab/flash-attention)

### Which has more tests, flash-attention or TensorRT-LLM?

TensorRT-LLM, with 2,703 test files; flash-attention has 80.

The proof: [NVIDIA/TensorRT-LLM's code at commit b433473](https://github.com/NVIDIA/TensorRT-LLM/tree/b433473ba0d0d49764f4d2fb3d45c82716b946dc)

### Which was updated more recently, flash-attention or TensorRT-LLM?

flash-attention: its last push was on November 16, 2026, and TensorRT-LLM's on November 8, 2026.

The proof: [Dao-AILab/flash-attention's commits on GitHub](https://github.com/Dao-AILab/flash-attention/commits)

### Which has more commits signed by AI, flash-attention or TensorRT-LLM?

In both, an AI signed 1.7% of their commits.

The proof: [the commits each tool signed in Dao-AILab/flash-attention](https://seerepo.com/r/Dao-AILab/flash-attention/ai)

## Sources

- Dao-AILab/flash-attention: Read on November 16, 2026, at commit 328dd7b (https://github.com/Dao-AILab/flash-attention/tree/328dd7b5a6905c75500a9bce334ac012763409a9)
- NVIDIA/TensorRT-LLM: Read on November 8, 2026, at commit b433473 (https://github.com/NVIDIA/TensorRT-LLM/tree/b433473ba0d0d49764f4d2fb3d45c82716b946dc)

## What is your repo built with?

Any public repo, with no account: https://seerepo.com/r/owner/repo/built, with the owner and name it has on GitHub.
