Skip to main content
Rase AI Intelligence Newsletter

Models & assistants Model or AI service

Aya 101 publishes a research model for instructions in 101 languages

Cohere’s Aya 101 is a multilingual research model trained to follow instructions in 101 languages. It adapts the mT5 architecture and is available as downloadable weights. [1][2]

Model access
Downloadable Aya 101 instruction-tuned research checkpoint. [2]
License
Apache License 2.0 for Aya 101’s published weights. [2]

What the model does

The team introduced its instruction-tuned model and multilingual evaluation in a February 2024 report. Its release documents training across both widely represented and lower-resource languages. [1][2]

Why it matters

One multilingual checkpoint lets researchers compare instruction following across languages with consistent model settings. That helps expose gaps that single-language testing would miss. [1][2]

Who can use it

Researchers developing multilingual evaluations can test the model on languages for which fewer labeled datasets or benchmarks exist. The model card links work on its training data. [1][2]

Key dates

  1. Aya 101 report first submitted [1]

Ways to use it

Use comparable prompts in two supported languages and have fluent readers check whether the content and quality stay consistent. Report results by language so weak performance is visible. [1]

Model limitations

Coverage of 101 languages does not imply uniform results across them. Aya 101 is the original research checkpoint and uses different model terms from later Aya Expanse releases. [1][2]

Sources and verification

Catalog evidence checked How we verify records

  1. Primary source Cohere research team’s original Aya 101 preprint and submission history
  2. Primary source CohereLabs/aya-101
  3. Other source Aya Model: An Instruction Finetuned Open-Access Multilingual Language Model

Related profiles

Models & assistants

GPT-6.2 Astra and Sol

OpenAI

Astra targets demanding projects. Sol targets everyday tasks. Both support text and image input in the API.

High confidenceEvidence checked

Models & assistants

MAI Code 2 and Code 2 Flash

Microsoft AI

Code 2 targets complex software work. Flash targets frequent edits, explanations and routine fixes.

High confidenceEvidence checked

Models & assistants

MAI Realtime 1

Microsoft AI

Users can try a MAI Labs demo. Developers can register for a limited Foundry preview.

High confidenceEvidence checked