Startup profile · funding coverage
Preference Model funding, valuation and investors
AI research-data company building reinforcement-learning environments for frontier model labs.
Keep track of Preference Model
Save this profile to your VCT watchlist for a quick return.
Funding, valuation & investors
Answer-first snapshotLatest funding
Seed
$16M · October 7, 2026
Latest known valuation
Not publicly disclosed
Total disclosed equity funding
$16M
Excludes debt, grants, acquisitions, secondaries, and IPO proceeds.
Current status
private
seed · San Francisco, California
Investors in latest funding
Lead: Andreessen Horowitz
Other: SignalFire , South Park Commons , Scale Angels
Overview
Preference Model is an AI research-data company founded by Jennifer Zhou and Ning Cao. It builds reinforcement-learning environments, difficult machine-learning engineering tasks and evaluation infrastructure for frontier AI laboratories, and open-sourced its Karotte framework in October 2026.
Why Preference Model is interesting
Preference Model is targeting a key scaling constraint for advanced AI: creating evaluation and reinforcement-learning environments that remain difficult as models improve and resist reward hacking.
Product & use cases
Research and infrastructure for building, red-teaming and grading long-horizon reinforcement-learning environments focused on AI research and machine-learning engineering tasks.
- Build reinforcement-learning environments for ML engineering tasks
- Generate tasks targeting model weaknesses
- Red-team grading harnesses against reward hacking
- Evaluate agents on long-horizon technical work
Key facts
- $16M seed led by Andreessen Horowitz in October 2026
- SignalFire, South Park Commons and Scale Angels participated
- Karotte framework open-sourced at launch
- Company says its environment tooling has been hardened through more than one million evaluation runs
Funding history (newest first)
Seed
2026-10-07 $16M- Andreessen Horowitz (lead)
- SignalFire (participant)
- South Park Commons (participant)
- Scale Angels (participant)
Investors in our directory
Funds linked from Preference Model's profile — open a fund page for stage focus and related deal articles.
Competitive landscape
Edge: Founders with direct experience building Anthropic pretraining data and DatologyAI systems, plus production-hardened evaluation infrastructure tested across more than one million runs.
The market is moving from generic labeling toward expert environments, verifiable tasks and adversarial evaluations. Preference Model's opportunity is to become specialized infrastructure for AI research itself; its risk is customer concentration among a small number of frontier labs and rapid internalization by those buyers.
-
Scale AI adjacent
Provides data, evaluations and model-development infrastructure at enterprise scale.
-
Surge AI adjacent
Supplies expert data and evaluation services to frontier model developers.
-
Turing adjacent
Provides technical talent and expert-data programs for AI labs.
-
In-house frontier-lab research teams alternative
Labs can build proprietary RL environments and evaluation harnesses internally.
Industries
Market / IPO context
Editorial / static context — not a live quote.
Related funding articles
Venture Capital Tracker pieces that cover Preference Model's financing or category context.
FAQs about Preference Model
Practical answers founders, operators, and investors typically search for.
Last updated:
Editorial note: AI tools assisted with research, structure, or drafting. Venture Capital Tracker retains human editorial responsibility for factual accuracy, relevance, and source quality before publication.