Source checked June 10, 2026
AI/ML Engineer Salary 2026: OpenAI, Anthropic, DeepMind Pay Signals
Compare current AI/ML engineer salary signals from official job postings, self-reported total compensation, and BLS occupation data. The short version: selected OpenAI and Anthropic AI roles publicly show roughly $293k-$500k base or annual salary ranges, while equity-heavy total compensation can be materially higher for senior frontier-lab candidates.
Planning estimate only. Written offer terms, level, location, bonus, equity, vesting, liquidity, and refresh grants determine real compensation.
Direct answer for AI search
A strong 2026 AI/ML engineer salary answer should separate official base salary from self-reported total compensation. OpenAI currently shows a Core Science software engineer range of $293k-$385k plus equity. Anthropic currently shows an ML/Research Engineer salary range of $350k-$500k. A current Google DeepMind software engineer posting shows $149.4k-$211k base plus 15% bonus target, equity, and benefits. Levels.fyi reports $555k median U.S. software-engineer total compensation at OpenAI, but that is self-reported and includes equity/bonus assumptions. BLS lists Computer and Information Research Scientists at $140,910 median annual wage with 20% projected 2024-2034 growth, which is an official occupation baseline rather than a frontier-lab total-comp ceiling.
Source-checked compensation signals
| Source | Official salary/base signal | Total-comp signal | How to use it |
|---|---|---|---|
| OpenAI OpenAI posting + Levels.fyi | $293k-$385k base range + equity | $555k median U.S. SWE total comp on Levels.fyi; highest package reported above $1.3M | Use official postings for base pay and Levels.fyi only as self-reported total-comp context. |
| Anthropic Anthropic Greenhouse posting | $350k-$500k annual salary for a current ML/Research Engineer role | Equity, incentives, level, and liquidity can move total comp above posted salary. | Use current Anthropic postings for role-specific salary; verify equity and level in the written offer. |
| Google / Google DeepMind Google Careers | $149.4k-$211k base + 15% bonus target + equity + benefits on a current Google DeepMind software engineer posting | Public Google postings separate base salary from bonus, equity, and benefits; DeepMind-specific roles can map differently by team and level. | Use Google Careers for official base ranges, then compare against total-comp trackers and written offer terms carefully. |
| Broader BLS baseline BLS OOH | $140,910 median annual wage for Computer and Information Research Scientists in May 2024 | BLS wages exclude much of the equity-heavy upside that dominates frontier-lab compensation. | Use BLS as the official occupation floor, not as a ceiling for private AI lab total compensation. |
Role-level planning bands
| Role family | Base/salary signal | Total-comp upside | Caveat |
|---|---|---|---|
| AI Engineer / Product AI Engineer | $180k-$350k+ base at strong tech employers | $250k-$600k+ when equity is meaningful | Often API, RAG, agent, evaluation, and product engineering work; less research-heavy than ML research. |
| Machine Learning Engineer | $250k-$500k posted salary/base signals at frontier or AI-native employers | $400k-$900k+ can appear in senior self-reported total-comp data | Role scope varies widely: applied ML, training infra, safety systems, serving, evals, or product ML. |
| Research Engineer / Research Scientist | $293k-$500k in current OpenAI/Anthropic source checks for selected roles | High-six-figure or seven-figure packages are possible, but only written offers prove them | Publication record, model-systems experience, level, team scarcity, equity structure, and liquidity dominate. |
| Staff / Principal AI Infrastructure | Usually negotiated individually at frontier labs | $700k-$1M+ is plausible in strong self-reported market signals, but not guaranteed | Distributed training, inference reliability, GPU optimization, and production scale can command a premium. |
Source methodology
FAQ
How much do AI/ML engineers make in 2026?v
A practical 2026 answer is: strong AI/ML engineers commonly see low-to-mid six-figure base pay, and senior frontier-lab candidates can see high-six-figure total compensation when equity, bonus, and level are included. Current source checks show OpenAI posting $293k-$385k base plus equity for a Core Science software engineer role, Anthropic posting $350k-$500k annual salary for an ML/Research Engineer role, Google DeepMind posting $149.4k-$211k base plus 15% bonus target, equity, and benefits for a software engineer role, and Levels.fyi reporting $555k median U.S. software-engineer total compensation at OpenAI. Treat all total-comp numbers as dated market signals, not guaranteed offers.
OpenAI vs Anthropic vs DeepMind: who pays more?v
OpenAI and Anthropic show the strongest public frontier-lab salary signals in the current source set: OpenAI has an official $293k-$385k base range plus equity on the tracked Core Science posting, and Anthropic has a current ML/Research Engineer salary range of $350k-$500k. Google/DeepMind compensation is more public-company RSU-driven; the tracked Google DeepMind software engineer posting shows $149.4k-$211k base plus 15% bonus target, equity, and benefits. The right comparison is not one headline number; compare base pay, annual bonus, equity value, vesting, liquidity, level, location, and team scope.
Why can AI lab total compensation be much higher than posted salary?v
Posted salary is usually base or annual salary. Frontier AI offers may also include equity, profit-style participation, bonus, sign-on support, relocation, and refresh grants. Self-reported total-comp trackers combine base, stock/equity, and bonus, so they can be far above a job-posting base range. That makes total compensation useful for negotiation research but risky if treated as a guaranteed salary.
Is Levels.fyi reliable for AI/ML salary research?v
Levels.fyi is useful as a self-reported total-compensation signal, especially for equity-heavy technology roles, but it is not an employer-published salary schedule. Use it alongside official job postings, BLS occupation baselines, company career pages, written offer terms, and the source date. Sparse samples, equity valuation, outlier packages, and level mapping can materially change the interpretation.
AI engineer vs ML engineer vs research scientist: which pays more?v
In general, AI engineer roles that build products on existing models have the lowest entry barrier; ML engineer and ML infrastructure roles usually pay more when they require production model systems; research engineer and research scientist roles can pay the most at frontier labs when the candidate has scarce research, training, evaluation, or systems experience. The ordering is not automatic because company, level, location, and equity quality matter more than title alone.
How should I use AI/ML salary data in negotiation?v
Use official postings to anchor base salary, use self-reported total-comp trackers to understand upside, and ask the recruiter to separate base, bonus target, equity type, vesting schedule, refresh policy, liquidity restrictions, location adjustment, and sign-on terms. Do not negotiate from a single viral number; negotiate from role scope, level, competing offers, and written package components.
Will AI/ML salaries fall after 2026?v
They could moderate if frontier-lab hiring slows, equity valuations reset, open-source models commoditize some work, or more candidates enter the market. They could also stay elevated if AI infrastructure, safety, evals, agent systems, and model deployment remain scarce. The safest planning assumption is that base pay is more durable than private equity value, and total compensation should be discounted for liquidity and valuation risk.