Find your job
// Quarterly report · Q2 2026

The state of contract hiring.

What's actually happening in the AI training and contract market this quarter. Volume shifts, pay movement, where work is moving, what's cooling. Updated every quarter.

// Headline

Math, medical, and code-RLHF are running away from the field.

+18%
Math reasoning pay · YoY
3.2×
Coding eval volume vs Q2 '25
$200/hr
Top medical domain rate
// Pay movement · Q2 2026 vs Q2 2025

Where pay is moving.

Median hourly rate movement across the 8 most-tracked roles. Year-over-year change.

Math reasoning · PhD
+18% UP
Domain · Medical
+15% UP
Coding eval (Rust/C++)
+14% UP
RLHF · Code reasoning
+11% UP
AI research · PhD
+8% UP
Coding eval (Python)
+5% UP
Multilingual (high-resource)
−4% DOWN
Creative · marketing copy
−7% DOWN

The hot end: anything that requires deep domain expertise or PhD-grade reasoning is up across the board. Frontier model training is increasingly bottlenecked by labelers who can actually verify hard outputs — and rates reflect that.

The cool end: high-resource multilingual annotation (Spanish, French, German) and short-form marketing copy are softening. The work hasn't dried up; the supply of qualified annotators has grown faster than demand.

// Volume shift

Where the jobs are.

Open roles in our index, by category, end of Q2 2026.

Coding eval
847
RLHF / Eval
412
Multilingual
389
Senior software eng
286
Math reasoning
276
Creative writing
234
Domain experts
187
AI research
94

Coding eval is twice the size of the next category. If you can credibly evaluate Python (or better, Rust / C++ / Go), this is the highest-supply / highest-demand work on the market right now — your time-to-first-paycheck is genuinely the fastest of any role here.

// Trend lines

Three things shifting under the surface.

1. The gap between top-tier and entry-tier rates is widening.

A senior coding evaluator now earns 2.4× a junior one — up from 2.0× a year ago. The same is true at the platform level: Mercor's senior tier vs entry tier is wider than it's ever been. Implication: tier-up early matters more than choosing the highest-paying platform on day one.

2. Domain-expert work is becoming credential-gated.

Six months ago, "domain expert" meant "subject-matter knowledge." Today, every major platform requires verifiable credentials — license, board cert, JD, MD, PE. Rates climbed accordingly. Implication: if you have credentials, lean in hard now; the rates aren't sticky and supply will catch up.

3. Multilingual is bifurcating.

Common languages (Spanish, French, German, Mandarin) are flat or down. Low-resource languages (Yoruba, Tamil, Vietnamese, Thai, Swahili) are up 25%+ YoY. Implication: if you speak a low-resource language fluently, this is the highest-leverage moment to enter the market.

// Predictions · Q3 2026

What we think happens next.

Code-RLHF rates climb again. Frontier model competition is shifting toward agentic tasks — labs that need labelers who can verify multi-step code reasoning will keep paying up. Expect $80–$110/hr to become the new senior-tier band on Outlier and Scale AI by end of Q3.

A new category appears: "tool-use" eval. As models get function-calling and browser-use abilities, evaluating those interactions becomes its own job. We expect 3–4 platforms to launch dedicated tool-use eval programs by September. Likely to debut at $60–$80/hr.

Domain-medical rates plateau. Active-practice MDs have been the highest-paid contractors on the market. Supply is finally responding; expect rates to flatten near $200/hr rather than continue climbing.

Multilingual high-resource keeps softening. The arbitrage from cheaper geographies has run its course; expect another −5–8% YoY for Spanish, French, German, and Mandarin annotators.

Ready to apply this?

Open roles updated every minute on joblet.ai.

Find your job Calculate your earnings →

// Methodology · Aggregated from public listings, ATS feeds, and 1,200+ contractor self-reports through Q2 2026.