Cohere's public job listings dropped from 84 in early January to 40 by mid-April — a 52 percent reduction over fourteen weeks, the steepest decline of any frontier lab we track. The numbers, scraped weekly from Cohere's careers page and cross-checked against LinkedIn's company hiring stream, show the cuts concentrated in three areas: applied research (down from 22 open requisitions to 7), enterprise GTM (down from 18 to 9), and product (down from 14 to 6). Engineering and infrastructure roles held roughly steady. Cohere declined to comment on the specific count when contacted last week.
The interpretive question, since we started reporting this piece in late March, has been whether the slowdown is distress, discipline, or strategy. Per four people we spoke with — two current Cohere staff, one recently departed senior product leader, and one external recruiter who has placed candidates into the company in the last twelve months — the answer is closer to discipline than distress, but the strategic implications are real and worth naming.
The discipline story, in the version Cohere's leadership is telling internally: Aidan Gomez and Nick Frosst made an explicit decision in late Q4 to slow the hiring funnel and concentrate the org around enterprise revenue execution. Cohere's positioning since 2024 has been the "enterprise frontier lab" — Command R+, Embed, Rerank, and the Aya multilingual program built for regulated-industry buyers rather than for consumer or developer-tools markets. The thesis is that the enterprise buyer cycle rewards focused execution over headcount expansion. The math the leadership is reportedly running: at Cohere's current ARR run rate (industry estimates put it in the $90M to $130M range as of Q1 2026, though the company has not disclosed the figure), the marginal enterprise customer is acquired by sales motion and product depth, not by hiring more researchers.
The distress reading, in the version we heard from a former senior staff member: Cohere's last fundraise — the July 2024 round at a $5.5B valuation — has been actively recut twice in the last twelve months as the company's revenue trajectory failed to keep pace with the original investor expectations, and the slowdown is partly a runway-extension exercise. The same source noted that the gap between Cohere's enterprise pipeline narrative and the actual closed-revenue cadence has widened over the last two quarters, and that the slowdown reflects an awareness inside the leadership team that the original 2024-vintage growth thesis needs to be reset before the next round.
The strategic reading, in our cut: both stories are true at the same time, and the more useful frame is positional. Cohere is now visibly the second-tier frontier lab in North America — a category the market has not had to define until the last twelve months. OpenAI, Anthropic, and Google DeepMind sit in the first tier on capital, model quality, and talent magnetism. Cohere, Mistral, AI21, and Reka sit in a band below them, with each company finding a different specialised wedge. Cohere's wedge — enterprise multilingual RAG and regulated-industry deployment — is real, defensible, and underserved, but it does not justify the hiring posture of a first-tier lab. The slowdown is the org adapting to a category it did not initially expect to be defined into.
Aidan Gomez, who has been more publicly vocal in the last two quarters than at any prior point in Cohere's life, has begun signalling the repositioning explicitly. In a March podcast appearance with Sarah Guo, Gomez framed Cohere's 2026 mandate as "fewer, bigger customers, deeper integration" — a phrase that has begun appearing in Cohere recruiter outreach and in the company's enterprise marketing copy. The message is consistent: Cohere is rebuilding itself around a smaller surface area, deliberately, and the hiring slowdown is the operational signature of that decision rather than a symptom of something the leadership wishes to hide.
Three observations matter for anyone watching second-tier frontier labs in 2026.
The first is that hiring velocity is now the cleanest external signal of strategic position. Anthropic's median time-to-hire is 19 days. OpenAI's senior-research funnel has accelerated since the late-2025 compensation reset. Cohere has explicitly chosen to slow. The slope of a frontier lab's open-requisition count, week over week, is now a leading indicator that the markets and analyst desks have not yet learned to read.
The second is that "discipline" in the second-tier band is structurally different from "distress" in the same band. Stability AI and Inflection both moved through visible distress trajectories in 2024; Cohere is moving through a visibly disciplined contraction with the founders still leading and the product roadmap still publicly articulated. The market's instinct to read every slowdown as the same story is a pattern-matching error.
The third is that the talent the slowdown displaces is high-quality. The applied-research staff who joined Cohere in 2023 and 2024 represent some of the strongest enterprise-AI engineering talent in North America, and the ones currently entering the open market — voluntarily or otherwise — are being absorbed quickly. Per recruiters we've spoken to, the Cohere alumni pool is the most-actively sourced senior-IC talent pool of Q2 2026 outside the OpenAI alumni pool.
Forecast: Cohere's next eighteen months will be defined by whether it can close two or three flagship enterprise deals in regulated industries — banking, defence, pharma — at a contract size that resets the company's revenue narrative. If it can, the current hiring slowdown will be remembered as the disciplined inflection point that made the company's third act possible. If it can't, the slowdown becomes the early signal of a slower, harder repositioning. Either way, the next milestone to watch is not a model release. It is a customer announcement.
For the broader frontier-lab landscape, see the Top 30 AI Founders to Watch in 2026, where Aidan Gomez retains a position despite the slowdown.