H Company (Runner H / Holo)

Paris agent lab whose Holo 3 claims to beat GPT-5.4 on cost

Agents & Automation Freemium Has API
Researched · Published · Reviewed
RECATOOLS Score
6.8 / 10
Capability
7
Value for money
6
Ease of use
6
ASEAN readiness
6
API quality
7
Founded
2023
HQ
Paris, France

Overview

H Company (Paris, founded 2023 by ex-DeepMind researchers, $220M seed) builds computer-use agents — Runner H (task automation), Surfer H (open-source browser agent), Tester H — powered by its Holo family of action models, now on Holo 3.

Advertisement

Pricing

Pricing shown for reference only. These figures reflect RECATOOLS research as of 3 Sep 2026 and may be out of date or incomplete. This is not financial or purchasing advice — always confirm the current price on the provider’s official website before making any decision.

Free
Free

What you can produce with H Company (Runner H / Holo)

  • Runner H autonomous browser-task agent (public beta)
  • Surfer H open-source browser-use agent (GitHub)
  • Tester H QA/testing agent
  • Holo 3 computer-use action model family
  • Open-weight Holo1 models and Surfer-H-CLI
  • Developer cookbook (Python/TypeScript agent SDKs)
Advertisement
RECATOOLS Verdict

H Company's research pedigree is real: the founding team came out of DeepMind, and the $220 million seed (Eric Schmidt, Amazon, Accel among the backers) was Europe's largest at the time. Holo 3's claimed 78.9% on OSWorld-Verified, ahead of GPT-5.4 and Claude Opus 4.6 on that specific benchmark at roughly a tenth of the inference cost, is a genuinely strong result if it holds up under independent testing — treat vendor-reported benchmark wins with the usual skepticism.

Runner H itself is still public beta: early reviewers report it losing context and getting stuck in navigation loops on complex multi-step tasks, alongside praise for how fast simple prompt-to-action tasks execute. Surfer-H-CLI and the Holo1 weights are genuinely open source on GitHub, which is unusual for a company this well-funded and worth crediting. Pricing isn't finalized — free beta now, enterprise quotes for heavier use — so budget-planning around this one is premature.

Independent AI-assisted assessment by RECATOOLS.

What people say

H Company doesn't yet have a G2 or Capterra presence with meaningful review volume — the product is young enough (Runner H beta launched November 2024, Holo 3 in 2026) that most feedback lives in Product Hunt comments, trade press, and social posts rather than structured review platforms.

What's there is a mixed but informative picture. On Product Hunt, reviewers call Runner H one of the stronger AI agents they've tried for straightforward web tasks — the prompt-to-action loop is fast, and it's positioned as approachable for people new to automation. The recurring complaint is reliability on harder tasks: reviews describe the agent compiling incomplete lists, mishandling language configuration, and in at least one documented case losing context mid-task and getting stuck in a navigation loop rather than recovering. That's consistent with a public-beta product rather than a finished one.

The more interesting signal is technical, via H Company's own published benchmarks and independent coverage from Sifted, Euronews, and FirstMark: Holo 3, released in 2026, reportedly scores around 78.9% on OSWorld-Verified, a benchmark for agents operating real computer interfaces, ahead of the figures H Company cites for GPT-5.4 and Claude Opus 4.6 on the same test, at roughly one-tenth the inference cost. Earlier-generation Holo-1 was benchmarked by the company at 5.5x cheaper than average competitors and hit 92.2% on WebVoyager when paired with the open-source Surfer-H-CLI agent, which the company put on GitHub along with a Python/TypeScript agent-building cookbook (hai-cookbook). Independent, non-vendor benchmark replication of the Holo 3 numbers hasn't surfaced yet in the coverage reviewed here, so the specific percentage claims should be read as company-reported rather than third-party verified.

Pricing remains unsettled: Runner H is free during beta, with H Company describing a future freemium-plus-enterprise structure but no published tiers, standing in contrast to competitors like OpenAI's Operator at $200/month for the Pro tier.

Summary of public user & expert reviews, compiled by RECATOOLS.

About this listing

Researched on
Published on
Last reviewed

This entry was compiled from publicly available data including H Company (Runner H / Holo)'s official website, press releases, documentation, and reputable third-party publications. RECATOOLS is not affiliated with H Company (Runner H / Holo) unless explicitly stated.

Data accuracy

Third-party AI tools update their pricing, features, availability, and policies frequently. Information here may be outdated by the time you read this — we make reasonable efforts to keep listings current, but cannot guarantee absolute accuracy.

For the latest details, please refer to H Company (Runner H / Holo) directly →

Spotted something out of date? Suggest an update →

Advertisement