Cohere for AI (C4AI)

Cohere for AI (commonly C4AI) is the non-profit open-research division of Cohere, the Canadian AI company headquartered in Toronto and San Francisco that is one of the principal commercial enterprise-AI vehicles globally. C4AI was established in 2022 by Cohere with a research-program mandate explicitly oriented around open-research output, multilingual foundation models, AI safety, and the broader research-community engagement that Cohere's commercial enterprise-AI offering does not directly serve. The division is led by Sara Hooker, the former Google Brain researcher and a senior figure in the broader open-AI research community. C4AI's flagship research output is the Aya multilingual-foundation-model program, which has released open-weights multilingual models trained on data across 101 languages with research-community engagement that has anchored substantive impact on multilingual NLP research globally. As of April 2026, Cohere for AI is one of the principal corporate-affiliated open-research divisions in the AI ecosystem.

At a glance

• Founded: 2022 by Cohere as a non-profit open-research division.
• Status: Non-profit open-research division of Cohere. Operates with research-program governance distinct from Cohere's commercial-enterprise organization.
• Funding: Operates within Cohere's research-and-development budget, with research-grant and academic-cooperation funding as supplementary support.
• CEO / Lead: Sara Hooker, Vice President of Research at Cohere and Head of Cohere for AI.
• Other notable leadership: Senior research scientists across the multilingual, AI safety, and academic-cooperation research-program areas.
• Open weights: Yes. C4AI's research outputs are released open-weights with full training-data and training-code disclosure where licensing permits.
• Flagship outputs: Aya multilingual-foundation-model program (Aya 23, Aya Expanse, and successors); active publication record at NeurIPS, ICML, ICLR, ACL, and other major AI and NLP venues; the C4AI Research Scholars program for researchers from under-represented regions.

Origins

Cohere for AI was established in 2022 by Cohere as a non-profit open-research division with a mandate explicitly oriented around open-research output and research-community engagement. Cohere's parent company, founded in 2019 by Aidan Gomez, Nick Frosst, and Ivan Zhang, had built a substantial commercial enterprise-AI business by 2022, and the C4AI division was structured to provide a research-community-aligned counterpart to Cohere's commercial offering.

The recruitment of Sara Hooker (former Google Brain researcher and a senior figure in the broader open-AI research community) as Vice President of Research at Cohere and Head of C4AI in 2022 anchored the division's research-credibility positioning. Hooker had been one of the principal voices in the academic-AI community for research on training efficiency, sparse models, and the broader applied-machine-learning research direction.

The 2022 to 2024 period saw C4AI build a substantive open-research program. The Aya project, launched as a community-driven multilingual-foundation-model effort with research collaborators globally, produced the Aya 101 model (covering 101 languages) and adjacent multilingual-foundation-model releases. The Aya project has been characterized in industry coverage as one of the principal multilingual-foundation-model open-research programs globally, with substantive impact on multilingual NLP research across academic and industry research peer communities.

The 2024 to 2026 period has continued the Aya cadence (Aya 23, Aya Expanse, and successor releases) alongside continued open-research output across NeurIPS, ICML, ICLR, ACL, and other major AI and NLP venues. The C4AI Research Scholars program has provided structured research-residency opportunities for researchers from under-represented regions, contributing to the broader academic-AI research-talent pipeline.

Mission and strategy

Cohere for AI's stated mission is to advance open-research and to build a research community that is more inclusive of researchers from under-represented regions. The strategy combines three threads. First, multilingual-foundation-model research with focus on the Aya program and broader multilingual NLP research areas. Second, AI safety research that contributes to the broader academic-AI safety research community. Third, the C4AI Research Scholars program and broader academic-community engagement that anchor the division's research-talent and research-community positioning.

The competitive premise is that an open-research division within a commercial AI company can produce research outputs (open-weights multilingual models, NLP research, research-community programs) that complement rather than directly compete with the commercial offering, while contributing structurally to the broader research-community ecosystem.

Models and products

• Aya multilingual-foundation-model line. Released open-weights with research-community engagement. Aya 101, Aya 23, Aya Expanse, and successor releases.
• Active publication record. Research at NeurIPS, ICML, ICLR, ACL, and other major AI and NLP venues. Research areas include multilingual NLP, AI safety, training efficiency, and applied machine learning.
• C4AI Research Scholars program. Structured research-residency opportunities for researchers from under-represented regions.
• Open-source contributions. Through the Cohere for AI GitHub organization and Hugging Face Hub.

Distribution channels include open-weights distribution through Hugging Face, open-source through GitHub, academic-publication through major AI and NLP venues, and the C4AI Research Scholars program.

Benchmarks and standing

Cohere for AI's evaluation framework focuses on academic publication metrics, multilingual NLP-benchmark performance of the Aya line, open-source-ecosystem adoption metrics, and the substantive impact of the C4AI Research Scholars program on academic-AI research-talent pipelines.

The Aya line has been characterized in industry coverage as one of the principal multilingual-foundation-model open-research programs globally, alongside BigScience BLOOM, HuggingFace multilingual research, AI4Bharat, Sarvam AI, Masakhane, and other multilingual-AI research peer organizations.

Leadership

As of April 2026, Cohere for AI's leadership includes:

• Sara Hooker, Vice President of Research at Cohere and Head of Cohere for AI.
• Senior research scientists across the multilingual, AI safety, and academic-cooperation research-program areas.

Funding and backers

Operates within Cohere's research-and-development budget. Cohere's parent company is privately held with substantive cumulative private capital exceeding $1 billion across multiple rounds.

Industry position

Cohere for AI occupies a distinctive position as one of the principal corporate-affiliated open-research divisions in the AI ecosystem, with the Aya multilingual-foundation-model line, the active academic-publication program, the C4AI Research Scholars program, and the substantive Hooker-anchored research-community engagement. Industry coverage has consistently characterized C4AI as one of the principal corporate-affiliated open-AI research divisions alongside Hugging Face, EleutherAI, and other open-research peer organizations.

Competitive landscape

• Hugging Face, EleutherAI, LAION, Allen Institute for AI (Ai2). Direct open-research peer organizations.
• BigScience, BigCode Project. Open-research-collaboration peer organizations.
• Cohere. Cohere for AI's parent commercial-AI company.
• Meta AI / FAIR, Google DeepMind, OpenAI, Anthropic. Frontier AI labs with adjacent open-research output.
• AI4Bharat, Sarvam AI, Masakhane. Multilingual-AI research peer organizations.

Outlook

• The continued Aya line cadence and multilingual-foundation-model open-research output through 2026 to 2027.
• The continued C4AI Research Scholars program and broader academic-community engagement.
• The continued research-publication output across NeurIPS, ICML, ICLR, ACL, and other major AI and NLP venues.
• The continued open-research-and-commercial relationship with the Cohere parent company.

Sources

• Cohere for AI. Research-program reference.
• Aya on Hugging Face. Open-weights multilingual-model line.
• Sara Hooker LinkedIn. Lead reference.
• Cohere. Parent company reference.
• Cohere for AI GitHub. Open-source contribution channel.