meta-llama/Llama-4-Scout-17B-16E-Instruct · Hugging Face

Fields

id

sources/llama-4-scout-hf

URL

https://huggingface.co/meta-llama/Llama-4-Scout-17B-16E-Instruct

Kind

repo

Accessed

2026-10-19

notes

Read https://huggingface.co/meta-llama/Llama-4-Scout-17B-16E-Instruct for the benchmark figures on models/llama-4-scout. As recorded by the reader: Values are from the official card's 'Instruction tuned models' benchmark table (Llama 4 Scout, 0-shot): MMLU-Pro 74.3, GPQA Diamond 57.2. MMLU-Pro used (not plain MMLU). Card also reports LiveCodeBench pass@1 32.8 but for the window 10/01/2024-02/01/2025, so it is not LiveCodeBench v6 and was omitted. Card also reports pre-trained-model MMLU 79.6, omitted as it is the base (non-instruct) model.

Notes

Read https://huggingface.co/meta-llama/Llama-4-Scout-17B-16E-Instruct for the benchmark figures on models/llama-4-scout. As recorded by the reader: Values are from the official card's 'Instruction tuned models' benchmark table (Llama 4 Scout, 0-shot): MMLU-Pro 74.3, GPQA Diamond 57.2. MMLU-Pro used (not plain MMLU). Card also reports LiveCodeBench pass@1 32.8 but for the window 10/01/2024-02/01/2025, so it is not LiveCodeBench v6 and was omitted. Card also reports pre-trained-model MMLU 79.6, omitted as it is the base (non-instruct) model.

Cross-references

This record references no other records.