This is the central comparison in AI right now — and the talent data reveals a more nuanced picture than the headlines suggest.

The headline: OpenAI is bigger (7,160 vs 4,568 — 57% larger), but Anthropic is growing faster (175% vs 80% YoY). Anthropic nearly tripled in 12 months; OpenAI didn't even double.

Hiring intensity favors Anthropic. In absolute terms, OpenAI hired more people (4,043 vs 3,341). But relative to base, Anthropic hired 73% of its current workforce in the last 12 months vs OpenAI's 56%. Anthropic is absorbing new talent at a significantly higher rate — a sign of both organizational confidence and execution capacity.

Hiring intent tells the same story. Anthropic has 717 open roles vs OpenAI's 706 — nearly identical in absolute terms. But as a percentage of headcount, Anthropic's 15.7% dwarfs OpenAI's 9.9%. Anthropic is planning to grow its workforce by a sixth; OpenAI by a tenth. For a company 57% smaller, that level of hiring ambition is notable.

Attrition is the other side of the coin. Anthropic's attrition rate is 10% vs OpenAI's 13%. That 3-point gap is meaningful at scale — it means OpenAI is churning through talent faster even as it tries to grow. Anthropic is both hiring at a higher intensity AND keeping people longer.

If you project these trajectories forward 12-18 months, Anthropic could close the headcount gap entirely. At 175% growth vs 80%, the lines cross sooner than most people expect.

They're fishing in overlapping ponds — but catching different fish.

The overlap: Google and Meta. Both companies pull heavily from Google (Anthropic 118, OpenAI 201) and Meta (Anthropic 97, OpenAI 207). These are the two biggest AI talent reservoirs on the planet, so the competition for Google/Meta alumni is direct and intense.

The divergence is where it gets interesting. Anthropic's #2 source is Stripe (103). OpenAI's is Apple (155). This tells you something about their strategies:
- Anthropic is pulling infrastructure and reliability engineers from Stripe — the people who build systems that handle billions of dollars without going down. That's an API-first, enterprise-reliability play.
- OpenAI is pulling product and consumer engineers from Apple and Meta — the people who build products 2 billion people use daily. That's a consumer-product play.

A critical note on OpenAI's Statsig inflow (106): OpenAI acquired Statsig — so these 106 "hires" represent an acquisition, not organic recruiting. Stripping out the Statsig acquisition, OpenAI's organic top-5 inflows would be Meta (207), Google (201), Apple (155), and AWS (68). The Statsig acquisition does signal investment in experimentation infrastructure (A/B testing, product iteration velocity), but it shouldn't be read as 106 individual engineers choosing OpenAI over alternatives.

The directional flow between them is the most telling data point. OpenAI lost 19 people to Anthropic in the last 12 months. Anthropic lost only 4 to OpenAI. That's a nearly 5:1 ratio. When people who've seen both options up close overwhelmingly choose Anthropic, that's a strong revealed-preference signal that recruiting spend alone can't replicate.

OpenAI has the edge on school pedigree — but employer pedigree tells a more nuanced story.

On raw school counts, OpenAI leads across the board: Stanford (443 vs 215), Berkeley (397 vs 184), MIT (233 vs 88), Carnegie Mellon (217 vs 82), Harvard (166 vs 87). OpenAI has roughly 2x the elite school alumni at each institution.

The more useful comparison is concentration — what percentage of each company comes from top-5 schools:
- OpenAI: 1,456 from top 5 out of 7,160 = 20.3%
- Anthropic: 656 from top 5 out of 4,568 = 14.4%

OpenAI has both more elite alumni AND a higher concentration. One in five OpenAI employees went to Stanford, Berkeley, MIT, CMU, or Harvard. For Anthropic, it's one in seven.

On employer pedigree, both companies are elite but with different flavors. Anthropic pulls from Google (118), Stripe (103), Meta (97), AWS (39), and Airbnb (29) — all five top sources are elite employers, a perfect score. OpenAI draws from Meta (207), Google (201), Apple (155), and AWS (68) — four elite employers in their top 5 (Statsig was an acquisition, not organic hiring). The difference: Anthropic's Stripe and Airbnb pipeline brings infrastructure and product engineering DNA. OpenAI's Apple pipeline brings consumer product and hardware engineering DNA. Both score near-perfect on employer quality.

Engineering concentration tells a different story. OpenAI is 41% engineering (2,961) vs Anthropic's 29% (1,321). OpenAI has built a deeply technical workforce — more than double the engineering headcount.

But Anthropic's lower engineering percentage reflects deliberate diversification. Anthropic has invested in Sales (6%), Administrative/Policy (6%), and Entrepreneurship/BD (5%). They're building a full-stack company with go-to-market muscle, not just a research lab. OpenAI's Operations at 14% and IT at 4% suggest a more infrastructure-heavy operating model.

The pedigree read: OpenAI has stronger school concentration. Both have near-perfect employer pedigree from different talent pools. If you're evaluating raw research firepower, OpenAI has the edge. If you're evaluating organizational readiness to commercialize, Anthropic's functional mix is more balanced.

Anthropic wins retention clearly. This is the most one-sided dimension in the entire comparison.

Start with attrition: Anthropic at 10% vs OpenAI at 13%. Three points doesn't sound like much, but at scale it's meaningful. OpenAI is churning through hundreds more employees per year, which means constant onboarding costs, knowledge loss, and cultural dilution.

The outflow numbers are where the gap becomes stark. Anthropic's top 5 departure destinations total just 22 people (Google DeepMind 6, Google 5, OpenAI 4, IBM 4, Apple 3). OpenAI's top 5 total 83 people (Meta 33, Anthropic 19, Google 15, Google DeepMind 8, Cursor 8). OpenAI is losing nearly 4x as many people to its top competitors.

The inflow-to-outflow ratio — the single best measure of talent magnetism:
- Anthropic: 386 in / 22 out = 17.5:1
- OpenAI: 737 in / 83 out = 8.9:1

Anthropic attracts nearly twice as many people per departure as OpenAI. Both ratios are strong (anything above 5:1 is healthy), but Anthropic is in a different tier.

The Meta dynamic is particularly telling. Meta is OpenAI's #1 departure destination (33 people). By contrast, Meta doesn't appear in Anthropic's top departure destinations at all — Anthropic's largest outflows are single digits (Google DeepMind: 6, Google: 5). Meta can pry people away from OpenAI, but not from Anthropic. When one of the highest-paying companies in tech can pull from your competitor but not from you, that's a retention signal worth paying attention to.

The direct rivalry flow reinforces this. OpenAI's #2 departure destination is Anthropic (19 people). OpenAI's #5 is Cursor (8 people) — a startup building on AI APIs. Meanwhile, Anthropic's losses to OpenAI are just 4 people. OpenAI is losing talent to both its primary competitor and to the ecosystem building on top of AI.

The skills data reveals fundamentally different bets on the future.

Top existing skills show the baseline DNA of each company:
- Both have Python as #1 (OpenAI 2,249 vs Anthropic 1,002) and Java in top 3. These are table stakes.
- OpenAI has C++ (1,399) and Algorithms (1,223) in its top 5. Anthropic doesn't. This points to OpenAI's deeper investment in low-level ML infrastructure, model training optimization, and research. C++ is how you squeeze performance out of GPU clusters.
- Anthropic has Data Analysis (637) and SQL (577) in its top 5 instead. This suggests a more data-driven, analytics-oriented workforce — useful for enterprise products, customer insights, and safety evaluation pipelines.

The fastest-growing skills are where the strategic divergence becomes clear.

OpenAI's growing skills are all product infrastructure: AWS (+117%), Elasticsearch (+116%), Cloud Applications (+116%), React.js (+113%), React Native (+112%). They're building a consumer product platform. The React Native growth indicates investment in mobile. This is a company building toward consumer-scale distribution — search, apps, daily-use AI.

Anthropic's growing skills tell a different story: Salesforce Implementation (+281%), TypeScript (+279%), Salesforce (+269%), Cloud IaaS (+252%). The Salesforce growth stands out. When an AI research company's fastest-growing skill is Salesforce implementation, that's a direct enterprise go-to-market signal. They're building the sales infrastructure to sell to Fortune 500.

The strategic read: OpenAI is building a consumer tech platform. Anthropic is building an enterprise AI company. These are different markets, different margin profiles, different growth curves — and potentially room for both to succeed at scale.

Anthropic scores 94/100 vs OpenAI's 84/100. The gap is meaningful — Anthropic leads on Retention (+4), Hiring Intent (+4), and Pedigree (+2). Both score perfectly on Acquisition and Skills Trajectory, reflecting the extraordinary talent magnetism of frontier AI labs. The Pedigree gap (20 vs 18) comes from school concentration: Anthropic's 14.4% elite school rate scores a perfect 10/10 on schools, while OpenAI's 20.3% does the same — but Anthropic gets a full 10/10 on employer quality (all five inflow sources are elite), while OpenAI's Statsig acquisition dilutes its organic employer score slightly.

Anthropic leads on trajectory. OpenAI leads on scale. The momentum favors Anthropic.

Here's the scorecard across every dimension:

OpenAI's advantages are real: 57% more employees, 41% engineering concentration (vs 29%), 20.3% elite school concentration (vs 14.4%), more C++ and Algorithms depth for core ML research. If you need the biggest bench of AI engineers today, OpenAI has it.

But Anthropic's advantages are the ones that compound:
- Growth: 175% vs 80%. Anthropic is closing the gap fast.
- Hiring intensity: Anthropic hired 73% of its current workforce in the last year vs OpenAI's 56%. And open roles represent 15.7% of Anthropic's headcount vs 9.9% for OpenAI. Anthropic is absorbing and planning to absorb talent at a proportionally higher rate.
- Retention: 10% vs 13% attrition. 17.5:1 vs 8.9:1 inflow-to-outflow ratio. Anthropic keeps its people at a measurably higher rate.
- Direct talent war: 19 people left OpenAI for Anthropic. Only 4 went the other way. That 4.75:1 ratio is the single most telling data point in this analysis. When engineers who've seen both companies from the inside overwhelmingly choose Anthropic, that's a revealed preference signal.
- Meta as a litmus test: Meta is OpenAI's #1 departure destination (33 people). Meta doesn't even appear in Anthropic's top departures. One of the highest-paying companies in tech can pull from OpenAI but not from Anthropic.

The strategic divergence matters: OpenAI is building a consumer tech platform (React, React Native, Elasticsearch). Anthropic is building an enterprise AI company (Salesforce, TypeScript, Cloud IaaS). These are different TAMs with different competitive dynamics. There may be room for both to be very large.

Talent signal: Anthropic has the edge on momentum. The 175% growth rate, the 73% hiring intensity, the 17.5:1 retention ratio, and the 19-to-4 directional flow from OpenAI are the kind of talent signals that tend to precede market position shifts by 2-3 years. OpenAI has the installed base, but Anthropic is winning the talent competition where it matters most — in the individual choices engineers make when they have both offers on the table.