Issue evidence · Serve · huggingface/text-generation-inference

Error when launching Llama 4 Scout

GitHub issue: https://github.com/huggingface/text-generation-inference/issues/3228

Not run

TGI 3.3.0 fails to launch Llama 4 Scout on 8x L40S 48 GB.

What Badgr ran

Nothing was run: Llama 4 Scout is a gated Meta model that needs an accepted license and Hugging Face token, and an 8x L40S node for a real run costs more than the $5 cap allows.

What came back

  • The reporter's environment is TGI 3.3.0, driver 560.35.03, CUDA 12.6.

A funded run on the same GPU topology with a token is the next step.

Checked 2026-10-14 in Badgr’s local development environment.

← All issue evidence