Sqwish Labs Raises $3M to Scale Decision Models
Sqwish Labs raised $3M in pre-seed funding led by Cambridge Enterprise Ventures to scale its AI decision models, outperforming larger competitors.

The big picture: Sqwish Labs, an eight-person AI startup based in Cambridge, UK, has raised $3 million in pre-seed funding. The company’s specialized AI decision models have demonstrated superior performance against systems from OpenAI, Microsoft, and TypeSafe in independent benchmark tests, suggesting that significant capital is not always a prerequisite for competitive AI development.
Why it matters:
- Agentic AI: As companies increasingly deploy AI agents that operate autonomously, the demand for fast, reliable decision-making systems is growing, often without the need for large language models to generate extensive text.
- Efficiency over Scale: Sqwish Labs’ approach challenges the prevailing industry trend of massive compute and funding, proposing that smaller, more focused models can handle routine decisions efficiently and cost-effectively.
- Market Validation: The startup’s strong benchmark results on platforms like JevBench and ImageJevBench provide crucial validation for its ‘kettei’ model family, positioning it as a viable alternative in a market dominated by tech giants.
How it works:
- Specialized Models: Sqwish Labs develops ‘kettei’ models, a family of AI systems designed to make rapid decisions and return structured answers with confidence scores, rather than generating long-form text.
- Benchmark Performance: The company’s largest model, kettei fat, ranked first among hosted decision models on JevBench and ImageJevBench, evaluating accuracy, confidence, speed, and cost for text and image-based decisions.
- Developer Platform: Sqwish offers its models through OneHot, a platform enabling developers to train customized decision systems. Additionally, kettei core and kettei dot are available as open weights on Hugging Face for on-premise deployment.
The catch: While Sqwish Labs’ benchmark performance is notable, translating these results into consistent, real-world reliability for enterprise customers remains the critical next step. The company’s claims of cost efficiency when paired with larger models are currently self-reported and require independent verification across diverse workloads. The decision model market is also attracting significant attention from major players, intensifying competition for market share and talent.
Key Facts
- Company: Sqwish Labs
- Amount: $3M
- Round: Pre-Seed
- Investors: Cambridge Enterprise Ventures (lead), Parkwalk, Innovate UK, Professor Steve Young, Karthik Tadinada
- Founders: Ushnish Sengupta, Federica Freddi
- Announced: 2026-11-17
- Sector: AI Decision Models
- Headquarters: Cambridge, UK
AIPressRoom Weekly
Every round we reported this week, the people behind the companies, and the events coming up. One email, every Tuesday.