A machine learning engineer in the US has a median total compensation of $282,950 on Levels.fyi (October 2026), against $196,500 for all software engineers. Software engineers who focus on ML or AI report a $250,000 median, about 27% more. Federal data is far lower because it counts wages and leaves out stock: the BLS May 2025 median for computer and information research scientists (SOC 15 1221) is $140,300, against $135,980 for software developers (SOC 15 1252). Comparing like for like, at the same company and level, Levels.fyi measured the AI premium at 6.2% for entry level and 18.7% for staff engineers in 2025.
Data as of October 8, 2026. Sources: Levels.fyi title, focus and city pages (updated October 7 and 8, 2026), Levels.fyi End of Year Pay Report 2025, BLS OEWS May 2025 via the Occupational Outlook Handbook, Dice Tech Salary Report (2024 survey), and posted pay ranges from OpenAI and Anthropic job listings. Updated October 2026.
How much does an AI engineer make in 2026?
It depends on which “AI engineer” you mean, and on whether you count stock. The best single benchmark is the machine learning engineer: Levels.fyi ML engineer data puts the US median total compensation at $282,950, based on 1,136 submissions, with a 25th percentile of $200,000 and a 90th percentile of $500,000.
The wider group of software engineers who list ML or AI as their focus has a median of $250,000 across 6,453 submissions, according to the Levels.fyi ML/AI focus page. That is the number to use when a hiring manager says “we need an AI engineer” but really means a strong software engineer who ships ML features.
The job title “AI Engineer” on its own tells a different story. Its median is $160,000 across 347 submissions on the Levels.fyi AI Engineer page. Many of those roles sit at companies building LLM applications on top of existing models, and they pay closer to general software rates than to ML specialist rates.
Which AI specialty pays the most?
Research scientists sit at the top of the US data, followed by ML engineers. Data scientists sit below the general software engineer median. The table shows every specialty we could verify, plus the ones where no reliable public median exists yet.
| Specialty | Median total comp (US) | 25th to 90th percentile | Sample |
|---|---|---|---|
| Research scientist | $310,715 | $182,500 to $553,500 | 1,003 |
| Machine learning engineer | $282,950 | $200,000 to $500,000 | 1,136 |
| Software engineer, ML/AI focus | $250,000 | $169,000 to $517,000 | 6,453 |
| Applied AI engineer (prompt and applied AI work) | $250,000 | $200,000 to $605,000 | 18 (thin) |
| Product manager (all, as AI PM proxy) | $230,000 | $170,000 to $465,000 | 5,384 |
| Software engineer (all) | $196,500 | $139,000 to $390,000 | 51,124 |
| Data scientist | $182,000 | $135,600 to $360,000 | 3,666 |
| AI engineer (job title) | $160,000 | $110,000 to $300,000 | 347 |
| MLOps, computer vision, NLP | No separate US median published by the sources we checked; price them off the ML engineer row | ||
Source: Levels.fyi pages for research scientists, applied AI engineers, product managers, software engineers and data scientists, updated October 7 and 8, 2026.
Why is “AI Engineer” lower than “ML Engineer”?
The two titles describe different labour markets. ML engineer is an established title at large tech firms that pay heavily in stock. AI engineer is a newer title, used widely by startups, consultancies and non tech employers building LLM features, where packages are mostly cash. The title alone is a weak signal, so price the req on the work, not the label.
What about MLOps, computer vision and NLP roles?
We could not find a separate, verifiable US median for these on Levels.fyi, BLS or Dice, so we do not print one. In practice they are ML engineer roles with a specific stack. Use the ML engineer range as the anchor, then move up for scarce skills such as GPU inference optimisation or production LLM evaluation, and down for roles that are mostly pipeline maintenance.
How does AI engineer pay change by level?
Pay roughly doubles from entry level to senior and triples by staff. The Levels.fyi 2025 pay report gives the US software engineer median total comp by level, and a separate Levels.fyi AI premium study measured how much more AI focused engineers earn at the same company and level in 2025.
| Level | All SWE median, 2025 | AI premium, same level, 2025 | AI engineer estimate |
|---|---|---|---|
| Entry | $155,000 | 6.2% | $164,610 |
| Mid (Engineer) | $226,000 | 11.9% | $252,894 |
| Senior | $312,000 | 14.2% | $356,304 |
| Staff | $457,000 | 18.7% | $542,459 |
| Principal | $551,000 | Not published | Not estimated |
The last column is a Work Insiders estimate. The arithmetic is the SWE median multiplied by one plus the premium, for example $312,000 x 1.142 = $356,304 at senior level. Levels.fyi measured the premium within company and level, so applying it to the overall median is an approximation, not a published figure.
Two trends stand out. The entry level premium shrank from 10.74% in 2024 to 6.2% in 2025, so junior AI roles are no longer much harder to price than junior software roles. The staff premium grew from 15.8% to 18.7%, which is where experienced AI talent is genuinely scarce.
What is the base salary versus total compensation for AI roles?
Total compensation on Levels.fyi is base salary plus annualised stock plus bonus. Base salary is what appears in job ads and in federal wage data. At large tech firms and AI labs, the gap between the two can be very large, because stock grants are a major part of the package.
Posted base ranges from employers in pay transparency states show the cash side. These are posted ranges for base salary only, as listed in October 2026, and they exclude equity and bonus.
| Employer and posted role | Location | Posted base range |
|---|---|---|
| Anthropic, Applied AI Engineer, Enterprise Tech | SF, NYC, Seattle | $200,000 to $320,000 |
| Anthropic, Applied AI Research Engineer | SF, NYC | $300,000 to $400,000 |
| Anthropic, ML Infrastructure Engineer, Safeguards Research | SF, NYC | $350,000 to $500,000 |
| OpenAI, Applied AI Engineer, Startups | San Francisco | $211,000 to $278,000 |
| OpenAI, Research Engineer (general) | San Francisco | $250,000 to $445,000 |
| OpenAI, Research Engineer, Codex | San Francisco | $380,000 to $500,000 |
| OpenAI, Product Manager, Core Models | San Francisco | $347,000 to $490,000 |
Sources: Anthropic listings for Applied AI Engineer, Applied AI Research and ML Infrastructure; OpenAI listings for Applied AI Engineer, Research Engineer, Codex Research Engineer and Core Models PM, all fetched October 8, 2026.
OpenAI states in its listings that total compensation also includes equity and performance bonuses on top of these ranges. Treat the posted base as the floor of a frontier lab offer, not the whole package.
Do frontier AI labs really pay close to $1 million?
For some engineers, yes, but they are outliers. The top three US software engineer employers by median total comp on the Levels.fyi software engineer page are Cursor at $975,000, Anthropic at $882,500 and OpenAI at $875,000. These medians come from self reported packages and include stock valued at the time of the offer, which at private labs depends heavily on recent funding valuations.
Those numbers describe a handful of companies. They are not a benchmark for a mid market AI req, and quoting them to a hiring manager outside that tier sets expectations no client can meet. We do not repeat individual “superstar” package reports here because we could not verify them from a primary source.
Which cities pay AI engineers the most?
Seattle and the San Francisco Bay Area lead for machine learning engineers, with medians above $325,000. New York is close to the national median. Austin and Boston come in lower, though their samples are small.
| Metro | ML engineer median total comp | Sample | All SWE median total comp |
|---|---|---|---|
| Greater Seattle | $335,000 | 152 | $259,480 |
| San Francisco Bay Area | $328,000 | 536 | $297,000 |
| New York City | $277,000 | 166 | $210,750 |
| Greater Los Angeles | $264,000 | 27 | Not in top five |
| Greater Austin | $218,500 | 32 | $212,800 |
| Greater Boston | $208,000 | 31 | Not in top five |
| US remote | No remote filter published in the Levels.fyi summary data we could verify | ||
Source: Levels.fyi ML engineer city pages for Seattle, the Bay Area, New York, Los Angeles, Austin and Boston, October 2026. Samples under 50 are directional only.
For remote roles, our working rule is to price against the employer’s pay band for its home market, since most large firms set remote pay by location tier. That is a Work Insiders recommendation, not a published figure.
How big is the AI pay premium over a regular software engineer?
Anywhere from 3% to 58%, depending on how you measure it. That spread is the most important thing for a recruiter to understand, because each figure answers a different question.
The raw median gaps (ML engineer 44% above all software engineers, research scientists 58% above) mix two effects: AI skills, and the fact that AI specialists cluster at high paying companies and senior levels. The same level comparison removes most of that mix and lands between 6.2% and 18.7%.
The Dice Tech Salary Report found professionals who design, develop or implement AI solutions earn 17.7% more than peers not involved in AI work. Dice notes that AI workers are also more likely to hold senior roles, which partly explains the gap. The survey covered 2,835 US tech professionals between August and November 2024, with an average tech salary of $112,521.
Why do AI salary sources disagree so much?
They measure different pay, different people and different years. Here is how the main sources compare.
| Source | Headline figure | Pay measured | Who is counted | Period |
|---|---|---|---|---|
| BLS OEWS, SOC 15 1221 | $140,300 median | Wages, no stock | Employer survey, 38,600 jobs | May 2025 |
| BLS OEWS, SOC 15 1252 | $135,980 median | Wages, no stock | Employer survey, all software developers | May 2025 |
| Levels.fyi ML engineer | $282,950 median | Base + stock + bonus | Self reported, skews to big tech | Rolling, Oct 2026 |
| Dice salary survey | 17.7% AI premium | Salary | 2,835 Dice users | 2024 |
| Posted job ranges | $200,000 to $500,000 base | Base only | Specific open roles | October 2026 |
The BLS figures come from the BLS research scientist profile and the BLS software developer profile. For SOC 15 1221 the lowest 10% earned under $82,200 and the top 10% over $230,630. For SOC 15 1252 the range runs from under $82,460 to over $214,670.
Federal wage data has no “AI engineer” occupation. Most ML engineers are coded as software developers, and AI researchers as computer and information research scientists. That is why the BLS gap between the two codes is only 3.2%, and why BLS is a floor for AI pay rather than a market rate.
Industry matters inside BLS too. Research scientists at software publishers had a median of $211,270 in May 2025, against $85,460 at state colleges and universities. Glassdoor and Indeed salary pages were blocked to our fetch, so we have left them out rather than quote them second hand.
What does this mean for recruiters pricing an AI req?
Start by classifying the role. Is it research, ML engineering, or software engineering that calls an LLM API? Those three pay very differently, and the title on the req often does not tell you which one it is.
Then pick the right benchmark for the client. For a big tech or well funded startup client, use Levels.fyi total comp at the right level. For a mid market or non tech employer paying mostly cash, anchor on the BLS 75th to 90th percentile and the “AI Engineer” title median of $160,000, then add the same level premium.
Quote base and total comp separately in every intake call. A candidate earning $250,000 total at a big tech firm may have a base near the low end of a posted $200,000 to $320,000 range. Losing a candidate because a client compared their total comp with your base offer is avoidable.
If you work AI reqs on a percentage fee, these numbers also change your economics. Run your margins through our cost per hire calculator, and check whether any AI screening tool you use is covered by local rules with our AI hiring law checker.
Where should recruiters source AI engineers?
Look beyond the obvious hubs. Seattle and the Bay Area pay the most, so candidates there are the most expensive to move. Austin, Boston and Los Angeles show ML engineer medians between $208,000 and $264,000, which gives clients outside the top two metros a realistic shot.
Strong software engineers who already ship LLM features are the largest untapped pool. With the entry level premium down to 6.2%, a client can hire and train a solid generalist for far less than it would cost to compete with frontier labs for a staff ML specialist. For wider demand trends, see our tech job market post and the labor market trackers.
How we built this page
We fetched every figure on this page from its source on October 8, 2026. Levels.fyi medians come from the structured summary pages for each title, focus and city, which are updated as submissions arrive. Level data comes from the Levels.fyi End of Year Pay Report 2025, and AI premiums from the Levels.fyi AI engineer compensation study published July 2025.
BLS wages are the May 2025 OEWS estimates published in the Occupational Outlook Handbook. Dice figures are from its salary report covering 2024 pay. Posted ranges were taken from live OpenAI and Anthropic listings. Any number labelled a Work Insiders estimate shows its arithmetic. For more benchmarks, see the Staffing Market Brief and our recruiting glossary.
Quick answers
What is the average AI engineer salary in 2026?
On Levels.fyi, the US median total comp is $282,950 for ML engineers, $250,000 for software engineers with an ML/AI focus and $160,000 for the job title “AI Engineer” (October 2026).
What does BLS say AI researchers earn?
The BLS median annual wage for computer and information research scientists (SOC 15 1221) was $140,300 in May 2025. That figure excludes stock.
How much more do AI engineers earn than software engineers?
At the same company and level, 6.2% more at entry level and 18.7% more at staff level in 2025, per Levels.fyi. Dice measured a 17.7% premium for AI workers in 2024.
Which city pays ML engineers the most?
Greater Seattle, with a $335,000 median total comp, just ahead of the San Francisco Bay Area at $328,000 (Levels.fyi, October 2026).
What do OpenAI and Anthropic post as salary ranges?
October 2026 listings show base ranges from $200,000 to $320,000 for an Anthropic applied AI engineer, and up to $500,000 for some OpenAI research engineers, before equity.
Do data scientists count as AI roles for pay?
Not on pay. The Levels.fyi data scientist median is $182,000, below the $196,500 median for all software engineers.
Why is the BLS number so much lower than Levels.fyi?
BLS counts wages only, from a survey of all employers, including universities and government. Levels.fyi adds stock and bonus and skews toward big tech.
Free newsletter
Get the Work Insiders weekly
One email a week: the recruiting and HR tech news that mattered, the tools we just added to the directory, and one practical guide. No spam, and you can unsubscribe in one click.