Skip to main contentThe Local Bench

Multilingual and translation · Cohere Labs

Aya Expanse 8B

Hugging Face: CohereLabs/aya-expanse-8b · checked 2026-10-13

8B version of Aya Expanse. Non-commercial licence.

Datasheet

Maker
Cohere Labs
Parameters
8.03 BVerifiedsafetensors.total from the HF API Source
Licence
CC-BY-NC-4.0VerifiedRestricts commercial use or is not declared. Read the licence before you build on it.
Access
Gated: accept the terms on Hugging Face to download
Task (HF pipeline)
text-generation
First published on HF
2024-10-23
Last updated on HF
2026-01-09
Weights in memory
16 GB at 16-bit · 4.9 GB at ~4.85 bitsEstimateParameters × bits ÷ 8. Excludes context cache and runtime overhead; not every model has a 4-bit build.

Benchmark

1,180
LMArena text, non-English category (Elo-style) · #3 of 3 scored in this category

c4ai-aya-expanse-8b, non_english, 95% CI 1170.9–1189.1, 4,321 votes. Source

The description summarises the model card at https://huggingface.co/CohereLabs/aya-expanse-8b. Parameters and licence are read from the Hugging Face API; the weights-in-memory line is our arithmetic, not a measurement.