Llama-4-Scout-17B-16E-Instruct config.json

Fields

id

sources/model-fr-hf-config-llama-4-scout

URL

https://huggingface.co/unsloth/Llama-4-Scout-17B-16E-Instruct/raw/main/config.json

Publisher

Hugging Face (unsloth mirror of Meta weights)

Kind

repo

Published

2025-04-05

Accessed

2026-10-03

notes

text_config: hidden 5120, 48 layers, 40 q heads, 8 kv heads, head_dim 128, 16 local experts, 1 per token, rope llama3 factor 16 from an 8192 base, max_position_embeddings 10485760. vision_config: 34-layer 1408-hidden encoder. | SOURCE AUDIT 2026-10-03: kind was 'spec-sheet' but the URL is a raw Hugging Face config.json served as text/plain, not a vendor spec sheet; set to 'repo'. The schema field it fills (layer/window/head counts) is machine-readable model metadata and should be read as such.

Notes

text_config: hidden 5120, 48 layers, 40 q heads, 8 kv heads, head_dim 128, 16 local experts, 1 per token, rope llama3 factor 16 from an 8192 base, max_position_embeddings 10485760. vision_config: 34-layer 1408-hidden encoder. | SOURCE AUDIT 2026-10-03: kind was 'spec-sheet' but the URL is a raw Hugging Face config.json served as text/plain, not a vendor spec sheet; set to 'repo'. The schema field it fills (layer/window/head counts) is machine-readable model metadata and should be read as such.

Cross-references

This record references no other records.