Working at Cohere in 2026: Transformer Pedigree, $7B Valuation, and Some Hard Truths

Cohere occupies a fascinating position in the AI landscape. Co-founded in 2019 by Aidan Gomez — a co-author of the landmark "Attention Is All You Need" paper that introduced the Transformer architecture — alongside Ivan Zhang and Nick Frosst, the company has grown into Canada's most prominent AI company. Valued at $7 billion with $1.54 billion in total funding, Cohere builds enterprise-focused large language models that compete directly with OpenAI and Anthropic for business customers.

But here's what makes Cohere genuinely interesting to analyze: it has some of the lowest employee satisfaction scores of any AI company we track. With a Glassdoor rating hovering around 2.9-3.2, work-life balance at 2.7/5, and only 44-48% of employees recommending the company, Cohere represents a case study in what happens when frontier AI ambition meets organizational growing pains. The technology is real. The compensation is strong. The internal experience is... complicated.

We pulled data from Cohere's company profile, employee reviews across multiple platforms, and industry analysis to give you an unvarnished picture of what it's actually like to work there in 2026.

Cohere at a Glance

Among the 118 companies in our Culture Directory, Cohere's Glassdoor rating of 2.9 places it near the bottom. For context, Anthropic sits at 4.1, OpenAI at 3.9, and even fast-growing AI labs like Mistral score 4.0. The gap between Cohere's external reputation (funded, credentialed, growing) and its internal reality (turbulent, politically complex, exhausting) is one of the largest we've seen in AI.

What Makes Cohere Different

Transformer DNA

Cohere's founding story is unusual. Aidan Gomez wasn't just inspired by Transformers — he helped invent them. As a co-author on the 2017 paper that changed everything in AI, he brings genuine research credibility that few AI startup founders can match. This matters because it attracts top-tier research talent who want to work alongside someone who was there at the beginning.

The research team at Cohere has published work on efficient models, retrieval-augmented generation, and multilingual AI that's well-regarded in the community. If you're an ML researcher who wants to publish, Cohere offers more academic freedom than most enterprise-focused AI companies.

Enterprise-first positioning

While OpenAI chases consumer attention with ChatGPT and Anthropic focuses on safety research, Cohere has deliberately positioned itself as the enterprise AI company. Their models are designed for deployment in corporate environments — data privacy, compliance, on-premise hosting, and sovereignty requirements that consumer-facing AI companies don't prioritize.

This strategic choice has real implications for engineers. You're not building chatbots for millions of users. You're building retrieval systems, generation pipelines, and embedding models that Fortune 500 companies run inside their own infrastructure. The problems are different: latency at enterprise scale, multi-tenant security, and compliance frameworks that constrain your architecture decisions.

Canadian talent advantage

With its headquarters in Toronto — home to the Vector Institute and a deep bench of ML talent — Cohere has access to a research pipeline that many SF-based companies envy. The company has expanded to San Francisco, London, New York, Montreal, Paris, and Seoul, but Toronto remains the center of gravity. Canadian compensation tends to be lower than Bay Area rates, which means Cohere can hire excellent researchers at costs that would be impossible in San Francisco.

Glassdoor Ratings: The Honest Picture

Cohere's Glassdoor profile is a red flag for anyone who filters by ratings. But the nuance matters. Let's break down what the numbers actually tell us.

The only sub-score above 3.0 is compensation (3.8), which confirms that Cohere pays well. Everything else — culture, management, work-life balance, career growth — scores below 3.0. This is unusual for a well-funded AI company and suggests systemic organizational issues rather than isolated complaints.

Important context: Cohere has grown from roughly 300 to 840 employees in two years. That kind of hypergrowth creates organizational stress that shows up in reviews. Multiple re-orgs, leadership turnover (they recently hired Meta's former AI head), and constantly shifting priorities are hallmarks of a company growing faster than its processes can keep up.

What Employees Actually Say

What employees love

The positive signals cluster around three themes: (1) the technology is genuinely frontier-level, (2) the people are brilliant, and (3) because the company is still maturing, there's enormous opportunity to own new domains. For engineers who thrive in ambiguity and want to shape rather than inherit systems, Cohere offers a blank canvas that more established companies cannot.

What employees warn about

The negative themes paint a clear picture: Cohere is a company under enormous competitive pressure that hasn't built the organizational infrastructure to handle its growth. The AI race creates urgency that conflicts with building sustainable engineering culture. When you're competing against OpenAI and Anthropic — companies with $10B+ in funding — the temptation to move fast and break things is constant. The result, according to employees, is re-orgs that feel reactive, communication gaps between leadership and ICs, and engineering processes that vary wildly between teams.

Compensation: The One Clear Strength

If there's one area where Cohere delivers unambiguously, it's compensation. The 3.8 rating for comp & benefits is the only above-average score in their Glassdoor profile, and the raw numbers confirm it.

Based on verified employee-reported compensation data, junior engineers (IC3, 0-3 years) earn approximately $255k in total comp ($170k base + equity + bonus). Mid-level engineers (IC4) jump to roughly $350k total ($200k base). Senior engineers (IC5) can reach $615k or more. These numbers are competitive with — and sometimes exceed — what Anthropic and OpenAI pay at equivalent levels.

The compensation structure includes base salary, equity (RSUs with standard 4-year vesting), and performance bonuses. AI Engineer roles command an 8-11% premium over general software engineering roles. For Canadian employees, the salary range is typically CAD $150k-$300k depending on level and location.

The implicit deal is clear: Cohere pays top-of-market because it knows the work environment is demanding. High comp partially compensates for the organizational turbulence and long hours. Whether that trade-off works for you depends entirely on your priorities.

Who Should Consider Cohere

Based on our analysis of the data, Cohere is an excellent fit for:

• ML researchers who want to publish — the research team has genuine academic freedom and Transformer pedigree attracts world-class collaborators.
• Engineers who thrive in chaos — if you find well-defined processes boring and want to help build them from scratch, Cohere offers that opportunity.
• Enterprise AI specialists — if you care about deployment, compliance, and making models work in production (not just on benchmarks), this is one of the best places to do that work.
• People optimizing for comp — at IC5+, Cohere's total comp rivals the best-paying AI labs in the world.

It's less ideal for:

• Engineers who need stability — if you want predictable priorities, clear career ladders, and well-established processes, Cohere's current stage will frustrate you.
• People who value work-life balance — the 2.7/5 WLB score is real. This is a company competing for its life against much better-funded rivals, and the hours reflect it.
• Mid-career engineers seeking mentorship — with career opportunities at 2.8/5, the growth infrastructure isn't there yet.
• Anyone allergic to re-orgs — multiple restructurings per year is the current norm.

Cohere vs. The AI Lab Landscape

How does Cohere stack up against the other AI labs you might be considering?

• vs. Anthropic — Higher Glassdoor (4.1 vs 2.9), stronger culture scores, safety-focused research. Anthropic is a better place to work by almost every metric, but is harder to get into and less enterprise-focused.
• vs. OpenAI — Similar intensity and political complexity, but OpenAI (3.9 Glassdoor) handles scale better. OpenAI offers more consumer-facing impact; Cohere offers more enterprise depth.
• vs. DeepMind — Completely different vibe. DeepMind is research-first, Google-backed, with better WLB and career infrastructure. Cohere offers more startup ownership and enterprise focus.
• vs. Mistral — Similar stage and ambition but Mistral (4.0 Glassdoor) appears to have healthier internal culture. Mistral is more open-source; Cohere is more enterprise-private.

The Bottom Line

Cohere in 2026 is a company of contradictions. The technology is genuinely frontier-level. The Transformer pedigree is real. The compensation is excellent. The enterprise positioning is smart and differentiated. But the internal experience — the re-orgs, the communication gaps, the immature processes, the work-life balance challenges — means this is emphatically not a company for everyone.

The 2.9 Glassdoor rating is a legitimate warning signal, not noise. It reflects a company growing faster than its organizational maturity can support, in an industry where competitive pressure makes sustainability feel like a luxury. Cohere is making a bet that top compensation and frontier AI work will attract talent willing to tolerate the turbulence. For some engineers, that bet pays off — particularly those who want ownership, research credibility, and enterprise-scale problems. For others, the same compensation and intellectual challenge are available at companies with healthier cultures.

The question to ask yourself: do you want to be part of building the organizational infrastructure that Cohere clearly needs, or do you want to work somewhere where it already exists? There's no wrong answer. But go in with eyes open.