We analyzed 71,061 salaried US job postings. Here's what pays the most in Q4 2026.

TL;DR. 71,061 US job postings from 8,405 companies disclosed a salary in the last 90 days. The median was $139,000. Rust postings paid a $216,000 median. Engineering postings that mention LLMs or GenAI paid $32,500 more than the ones that don't. Salesforce and SQL sit at the floor around $130-140K.

The dataset

Our scanner pulls US job postings off company careers pages — Greenhouse, Lever, Ashby, Workday, SmartRecruiters, Recruitee, Teamtailor, plus curated ATS feeds. 145,074 US postings came through between Aug 1 and Oct 30. 71,061 of them disclosed a salary we could parse into a sane annual number (both ends present, both between $20K and $1M). That is the set for this post.

• Postings: 71,061

• Companies: 8,405

• Window: Aug 1 – Oct 30, 2026

• Median comp (midpoint of range): $139,000

• 25th: $100,000

• 75th: $185,500

• 90th: $240,000

Methodology at the bottom.

AI mention inside engineering is worth $32,500

Same role domain. Same window. Same salary filter. The only difference is whether the posting talks about LLMs, generative AI, OpenAI, Anthropic, Claude, ChatGPT, or RAG.

21.7% delta on the median. 18% of engineering postings in the window mention AI at least once. That share has been climbing every week we have measured it.

Skills ranked by what they pay

Regex over the lowercased title + description + requirements for 23 skills and tools. Word boundaries where it mattered (Rust, Go, Python, Spark). A posting counts once per skill, no matter how many times it says the word. US only, salary disclosed, last 90 days, 100+ postings minimum to show up.

Rust at the top is a scarcity number. 753 postings is a thin slice of the market, and most of the Rust shops hiring right now are infra, protocols, databases, or AI runtimes. Those shops are also the best funded. The CUDA line is the same story in a different vertical — if you can write a kernel, somebody at an AI lab is overpaying for your time.

Azure trails AWS and GCP by $8-11K across the sample. Consistent, not huge.

SQL and Salesforce are the floor. Both show up in roughly 11% of all US salaried postings — enormous surface area. The roles behind them are business analyst, salesops, admin, revenue ops. Those roles pay below the engineering median.

By role domain

Every scraped posting lands in one of 14 role domains. The ones with enough US salaried postings to compare:

PM clears engineering by $15K at the median. The PM IQR is tight though — $133K to $215K. There is no $500K PM job in the sample. There are plenty of $500K staff eng jobs. The 50th percentile goes to PM. The 95th percentile goes to engineering.

By specific job title

Keyed off the title text, not the role domain classifier:

Staff Engineer is the first title where the median breaks $220K. Engineering Manager and Principal both clear $200K. The AI/ML Engineer number is $195K across 743 postings, and that is before the top labs, which show up next.

Highest-paying companies

Companies with 20+ US salaried postings in the window:

The top two are composition. Radiology Partners posts a lot of radiologists, which clear $500K before you even talk about equity. Netflix publishes very wide bands that frequently top $700K at the senior end. Both inflate the median compared to a software shop.

The middle of the table is the real signal. Anthropic at $335K across 485 postings is not a rounding error. That is a sustained offer level across a real hiring footprint. NVIDIA at $270K across 364 is the same picture. Figma, Databricks, Reddit, Square. The $300K-ish bar is not a FAANG-only phenomenon anymore.

By state

Postings tied to a US state, not remote. 100+ postings in the window.

Washington beats California at the median. Composition — WA onsite is heavily Microsoft, Amazon, and Meta Seattle. California includes a broader mix of mid-market SF and LA companies that pay less than the Seattle giants.

DC is a surprise at $249K. Thin at 110 postings. Defense contractors and biotech both pay well for cleared or specialized roles, and both are heavy in the DC sample.

Remote vs onsite

62,647 remote US postings. 8,414 onsite. Both disclosed salary.

• Remote: $133,129

• Onsite: $191,295

$58K gap at the median. That is not role-for-role. The onsite population and the remote population are not the same jobs. Companies that still ask people to come in are the companies that pay the most to put them in seats — staff-plus roles at FAANG-adjacent HQs, specialized research and legal, senior engineers at headquartered companies. The baseline remote posting is a senior IC at a Series B-D paying $130-140K. Different populations, not different pay for the same work.

Hybrid is barely a labeled category here. Most companies tag "remote" or do not tag at all. If your calendar is three days in office, the ATS almost certainly says remote.

About the research

Data source. All postings were collected by the HiringFunnel scanner between Aug 1 and Oct 30, 2026 from company careers pages (Greenhouse, Lever, Ashby, Workday, SmartRecruiters, Recruitee, Teamtailor) and curated ATS feeds. We re-fetch continuously. Stale postings are reaped.

Salary parsing. Three sources: explicit ATS fields (83,606 postings in the salary set overall), extraction from job description text (31,327), and language-model inference where neither of those two is present (7,283). For this post we only used postings where salary_min and salary_max were both populated and both sat between $20,000 and $1,000,000 annual. That removes hourly rates misread as annual, placeholder zeros, and mega-outliers from scraper bugs (the worst one in the raw data had a max of $4.5 trillion, which is impressive).

Role classification. role_domain and primary_tech_stack are enum fields populated by a language-model classifier that reads the title and description. The specific-title breakdown (Staff Engineer, AI / ML Engineer, and so on) is a regex pass over the title text.

Skill matching. Each skill is a regex over the lowercased title + description + requirements, with word boundaries where ambiguous (Rust, Go, Python, Spark). One count per skill per posting, no matter how many times it appears.

What is not in the set. Non-US postings. Postings older than 90 days. Postings without both ends of a salary range. Postings where either end sat outside the $20K-$1M window. Everything else is in.

Want the raw query, a different cut, or the data behind one of these tables? Email me.

If you are an engineer looking for a job and you want this kind of filtering run on your own search — salary floor, remote preference, countries that work for you, skills that pay — the application is open.