---
title: "Vals AI Raises $40M Series A at $400M — a16z Leads Independent AI Evaluation"
description: "Vals AI announced a $40M Series A at $400M on August 13, 2026, led by a16z, with 8VC, Pear VC, and BloombergBeta returning. The product is independent benchmarks for real work — not another model lab."
date: 2026-08-13T00:00:00.000Z
tags: ["2026-vc-news", "startup-funding", "venture-capital", "artificial-intelligence", "a16z"]
source: https://venturecapitaltracker.com/2026-vals-ai-40m-series-a-a16z
---

# Vals AI Raises $40M Series A at $400M — a16z Leads Independent AI Evaluation

> Vals AI announced a $40M Series A at $400M on August 13, 2026, led by a16z, with 8VC, Pear VC, and BloombergBeta returning. The product is independent benchmarks for real work — not another model lab.

**Vals AI** announced a **$40 million Series A at a $400 million valuation** on **August 13, 2026**, led by **[Andreessen Horowitz](/fund/andreessen-horowitz)**. Returning names: **[8VC](/fund/8vc)**, Pear VC, BloombergBeta. New: HRT Ventures and Next Ladder Ventures.

The company is **not** training a frontier model. It is selling **independent measurement** of models that already exist.

## Key facts

| Field | Detail |
| --- | --- |
| Company | Vals AI — independent AI evaluation (San Francisco) |
| Round | $40M Series A @ $400M |
| Date | August 13, 2026 (company blog) |
| Lead | Andreessen Horowitz |
| Returning | 8VC, Pear VC, BloombergBeta |
| New | HRT Ventures, Next Ladder Ventures |
| Traction (company) | Revenue 8x vs all of 2025; customers 2x; team 3x in six months |
| Releases with the round | Vals Smith (GA), frontier-risk benchmarks, Vals Index 2.0 |

## Who uses the product — and for what job

**Users:** model labs that need a citation that is not their own eval harness; enterprises picking models for finance/legal/coding work; governments measuring capability and cyber risk.

**Job:** build a **hill that is not in the training set** — then score models on whether they can do **paid professional work**, not trivia.

Company: results cited in model cards from OpenAI, Anthropic, Google, Meta, and xAI; work with the U.S. Department of Commerce and members of Congress (Vals post — treat as company-reported).

## Why now

- Public benchmarks **saturate and leak**. Vals’s thesis: the industry spent trillions on intelligence and comparatively little on **independent hills**.
- Enterprises are being asked to **spend on AI without ROI measurement**.
- Same week as [Databricks’s $190B close](/2026-databricks-5b-190b-coatue) and [Cursor’s SpaceX close](/2026-spacex-closes-cursor-acquisition): capital is piling into **builders**. Vals is a **scorekeeper** round.

## Why a16z / 8VC — portfolio fit

**[a16z](/fund/andreessen-horowitz)** leading a **$400M** eval company is a **picks-and-shovels** check next to lab investments. The conflict to underwrite: can an a16z-backed evaluator stay **credible** with labs a16z also funds?

**[8VC](/fund/8vc)** returning from seed is continuity on a **national-capability** theme (measurement as infrastructure).

**Likely founder rationale:** take the brand that already sits in model-card footnotes and fund **private test-set** expansion (Vals Smith, risk benches).

| Investor type | What they bring |
| --- | --- |
| a16z | Lead; lab/enterprise network; category signal |
| 8VC | Seed continuity; defense/infra pattern recognition |
| HRT / Next Ladder | New capital; HRT adjacency to quantitative evaluation culture |

## Competitive map

| Player | Lane |
| --- | --- |
| Vals AI | Private professional-work benches + enterprise eval platform |
| LMSYS / arena-style leaderboards | Public preference voting |
| HELM / academic suites | Research transparency; slower productization |
| Lab-owned evals | First-party; conflicted when used as marketing |
| Red-team / safety vendors | Overlap on cyber/mental-health risk benches Vals is adding |

## Market signal

**$40M at $400M** with **company-reported 8x revenue** says investors will pay **growth-equity prices** for a **measurement** company if labs already cite it. It is still **tiny** next to [Anthropic’s $65B Series H](/2026-anthropic-65b-series-h-965b-valuation). That is the point: **scorekeepers raise tens of millions; labs raise tens of billions.**

## Caveats

- Vals is an evaluation company, not a foundation-model lab.
- The 8x revenue figure is **company-reported**, not audited.
- a16z also funds labs that Vals scores — independence vs portfolio overlap is the diligence question.

## Practical takeaway

- **Founders (AI eval):** Sell **private tests + buyer job** (procurement, policy), not another public leaderboard.
- **Investors:** Underwrite whether citations in model cards convert to **renewable enterprise contracts**.
- **Operators:** Use Vals (or peers) when the question is **“can this model do our work?”** not **“who won the arena?”**

### Sources

1. Vals AI (Aug 13, 2026): https://www.vals.ai/blogs/series-a
2. Related: [/startup/vals-ai](/startup/vals-ai) · [/fund/andreessen-horowitz](/fund/andreessen-horowitz) · [/fund/8vc](/fund/8vc) · [/2026-august-14-18-investment-news-ai-infra-physical](/2026-august-14-18-investment-news-ai-infra-physical)