Skip to content
← Models
Open weightsCohere For AI/10 trustMetrics-checked

Aya 23

Open-weight multilingual models covering 23 languages in depth.

Visit official page →

About

Aya 23 was released on May 23, 2024 by Cohere For AI, the non-profit research arm of Cohere, in 8B and 35B parameter variants. Where the earlier Aya 101 chased breadth across 101 languages, Aya 23 goes for depth: a highly performant pre-trained model paired with the Aya dataset collection, serving 23 languages spoken by nearly half the world's population.

The 35B variant builds on a refined Command R, and both models outperformed Aya 101, Gemma, and Mistral-class models on multilingual discriminative and generative benchmarks at release. Weights are public under CC-BY-NC 4.0 — a non-commercial license, so commercial use requires separate permission.

Specs

Context window

Pricing

Free quota

API access

Weights

Open weights

Released

May 23, 2024

Benchmarks

Provider-reported unless a source says otherwise. Benchmarks are marketing-adjacent by nature — treat them as directional.

Use cases

multilingual-chattranslationsummarization

Trust score — 7/10

/10 trustMetrics-checked

Scored 7/10 (metrics tier, metrics-v1). Evidence: Verified via vendor site/docs/GitHub and reputable coverage, wave 74; Trending score 70/100 from researcher assessment.

More in the aya family

Sources