Stockmark
Tokyo AI for manufacturers' R&D, with its own Japanese LLMs
Overview
Stockmark is a Tokyo company building AI for Japanese manufacturers: Aconnect (formerly Anews), an R&D agent platform for news, patents and papers, and SAT Agent Cockpit for structuring company data, plus its own Stockmark-2 LLMs. Pricing is quote-only, with free trials on request.
Pricing
Pricing shown for reference only. These figures reflect RECATOOLS research as of 1 Oct 2026 and may be out of date or incomplete.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.
- R&D research and patent-search agents
- Free trial unlimited in users and features
- Structuring of company documents
- API/MCP connection to existing agents
- Trial limited in period and users
What you can produce with Stockmark
- Aconnect (formerly Anews) — R&D agent platform for news, patents and papers
- SAT Agent Cockpit — structures company data for existing AI agents via API/MCP
- Stockmark-2-100B Japanese/English LLM (from scratch), offered as NVIDIA NIM
- Nemotron-based JapanDocReader open model with NVIDIA
- BusinessSlideVQA benchmark for Japanese business documents
- ISO 27001, 27017 and 27701 certification
Stockmark is a focused Japanese enterprise AI vendor: Aconnect gives manufacturing R&D teams research agents over news, patents and journals, and SAT Agent Cockpit structures internal documents for the AI agents a company already runs. It backs that with its own models, including Stockmark-2-100B and a document-reading model built with NVIDIA, and with strong ties to Japanese industry — GENIAC selection, AWS and NVIDIA partnerships, and an AI-Ready project with Suzuki, SMBC and others. Older descriptions quoting a free tier and ¥29,800 plans are from 2017; everything is now quoted, with trials on request. Its claimed reach into 30% of the Nikkei 225 comes from its own materials. A strong option for large Japanese manufacturers; outside Japan it has little presence.
What we found
Stockmark is a Tokyo company, founded in November 2016 and led by chief executive Tatsuya Hayashi, that builds AI for Japanese manufacturers' research and strategy teams. It runs two businesses. Aconnect, renamed from Anews in July 2025, is an AI agent platform for manufacturing R&D: it gathers news and research from about 35,000 sites, patents and journal papers, and adds agents for idea generation, market-size estimation, technology search and patent search; customers on its site include Resonac, Asahi Beer, Lion, Taiyo Yuden and ORIX Bank. SAT, now "SAT Agent Cockpit", structures a company's own documents and connects to its existing AI agents through APIs and MCP, with case studies from Mitsubishi Chemical, JTEKT, Teijin, NGK and Itochu. An older strategy tool, Astrategy, still has a page but has left the main navigation.
It also trains its own models. Stockmark-2-100B was built from scratch on about two trillion tokens — 60% English, 30% Japanese and 10% code — and is offered as an NVIDIA NIM microservice; a 100-billion-parameter vision-language model remains in beta under the Qwen licence because it was trained partly on Qwen-generated data. In July 2026 it released, with NVIDIA, a commercially usable open model for reading Japanese documents built on Nemotron, and it has a strategic collaboration with AWS. It has been selected for every round of Japan's GENIAC programme and holds ISO 27001, 27017 and 27701.
Stockmark raised about ¥4.5 billion in a Series D led by Polaris Capital in October 2024, taking its total above ¥8.8 billion. In May 2026 it launched an "AI-Ready" project with 16 major companies including Suzuki, Yanmar, Mitsubishi Chemical, SMBC and Kobe Steel. Its own 2024 materials said 30% of Nikkei 225 companies and more than 300 firms used its products, 70% of them manufacturers.
No prices are published: both products are quoted from an initial fee, a base fee with per-user licences and options, and free trials are available on request. It suits large Japanese manufacturers; we found no Southeast Asian presence.
Compiled by RECATOOLS from public sources: vendor documentation and pricing, code repositories and published reporting.
About this listing
This entry was compiled from publicly available data including Stockmark's official website, press releases, documentation, and reputable third-party publications. RECATOOLS is not affiliated with Stockmark unless explicitly stated.
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 Stockmark directly →
Spotted something out of date? Suggest an update →
Alternatives to Stockmark
More in Research & Data
