Genspark Details Gen-1 Slides Model and Enterprise Rollouts

Genspark outlined its proprietary AI model development and enterprise agent strategy, including the September 18 launch of Gen-1 Slides, a model specialized for slide generation, and company-wide deployments at Leopalace21 and Nisso Corporation, according to watch.impress.co.jp.

The company says Gen-1 Slides reaches quality comparable to Opus 5 in public benchmarks and internal evaluations at about 6% of the cost of the comparison model. Genspark also described cost controls such as model routing, which assigns low-cost models to simple questions and high-performance models to complex reasoning, and Advisor Mode, in which a low-cost model handles normal processing and consults a high-performance model when work stalls. Paid plan users receive a weekly usage allowance usable without consuming credits.

On the development side, Genspark says it employs 49 software engineers who built 28 products and features over the past year. Its Lights-off Factory lets multiple AI agents handle implementation, testing and code review of GitHub tasks once a human selects them, with humans making the final call on merging code. A single engineer developed the open-source office software GenOffice in about one week using roughly 10,000 dollars' worth of AI usage credits.

Leopalace21 is rolling Genspark out to about 4,000 officers and employees. The company set up a DX promotion department in April, clarified what information may and may not be entered into AI, and gave accounts to employees who passed training and a confirmation test. It began the company-wide rollout in July, built an environment where common templates can be used for document creation, and also uses Genspark for disaster response such as organizing and analyzing damage information.

Nisso Corporation, a group company of NISSO HOLDINGS, built AI agents for recruiting operations. It unified rules across its 13 group companies over about two months and selected about 60 people, mainly in their 20s and 30s, for the initial deployment. Its AI employees cover three tasks: job market analysis, job posting draft support, and matching job postings with applicants.

Genspark said it uses US-based inference infrastructure for open-weight models and has Zero Data Retention contracts with external AI model providers so input data is not used for model training.