This article covers Sqwish Labs, a Cambridge-based AI startup, which has raised £2.3m in a pre-seed funding round led by Cambridge Enterprise Ventures with participation from Parkwalk and co-funding from Innovate UK. The funding will back further development of kettei, its fast decision models that assign probabilities to structured outputs and are intended to reduce reliance on larger, costlier models, supporting startups and organisations deploying structured decision-making in AI systems.
Cambridge-based Sqwish Labs has raised £2.3m in a pre-seed funding round led by Cambridge Enterprise Ventures, the University of Cambridge’s investment arm, with participation from Parkwalk and support from Innovate UK. The funding will bankroll further development of kettei, Sqwish’s fast decision models that assign probabilities to structured outputs and aim to reduce reliance on larger, costlier models.
Sqwish’s announcement arrives at a moment of intense interest in “decision models” — compact AI systems tuned to produce structured answers (for example, yes/no responses, multiple-choice selections or numerical scores) rather than free-form text. These models are likely to be embedded in agentic AI systems that must decide when to act, which tool to call and when to escalate to a larger model. If Sqwish’s claims hold up in production, the company’s approach could materially lower cost and latency for many routine decisions in AI stacks.
Sqwish builds kettei models that return an answer and an associated probability in milliseconds, rather than generating text. Its largest model, kettei fat, launched on 14 October and, as of 29 October, ranked first for capability on JevBench — an independent decision-model benchmark — ahead of offerings from OpenAI, Microsoft and TypeSafe. On ImageJevBench, which evaluates decisions about images, kettei fat placed first out of 64 systems, and it sits second on S1MB’s English leaderboard, ahead of Perplexity and Cloudflare.
The company also offers two open-weight models, kettei core and kettei dot, available for download on Hugging Face so teams can run them in-house and keep data on their own systems. Sqwish’s OneHot platform lets customers upload examples and produce a bespoke model in minutes, with interfaces including Python and TypeScript SDKs, an MCP server and a Claude Code plugin. The models are designed to be a cheap front line: Sqwish quotes pricing at $0.02 per million input tokens, meaning around three US cents for a thousand 500-token decisions on kettei fat.
Sqwish says its first products focused on compressing and routing requests to larger models; earlier in the summer its model router topped RouterArena, an open leaderboard for routers. The current roadmap prioritises expanding kettei’s outputs beyond choice probabilities to include calibrated numbers, dates and short passages of text, multilingual support and continual learning from real-world feedback — research that Innovate UK is co-funding.
The round was led by Cambridge Enterprise Ventures, the University of Cambridge’s investment vehicle, with participation from Parkwalk and co-funding support from Innovate UK. New angel backers include Professor Steve Young, Emeritus Professor of Information Engineering at Cambridge and founder of VocalIQ (acquired by Apple and integrated into Siri); Karthik Tadinada, CEO of Fortify AI and former COO of fraud-detection firm Featurespace; and partners from Highland Europe investing in a personal capacity.
Earlier supporters converted into this round: venture firm Vento; Mandeep Singh, former CEO of online marketplace Trouva; Jon Bradford, partner at Dynamo Ventures; and Guy Fighel, co-founder of SignifAI (acquired by New Relic). Innovate UK’s involvement is specifically tied to the research programme behind kettei and the next-generation models Sqwish plans to build.
The investor mix combines university-backed venture capital, specialist funds and experienced operator angels, signalling interest from both academic-transfer vehicles and AI-focused investors in compact, decision-oriented models.
If you're researching potential backers in this space:
In the announcement, Ushnish Sengupta, co-founder and CEO at Sqwish Labs, said:
AI doesn’t need a giant model, or a bigger data centre, for every small decision. A small, fast model can make most of them for a fraction of the cost and energy, and knows when to ask for help. This funding takes kettei from the top of the leaderboards into the products people use every day.
Sqwish plans to use the funding to train the next generation of kettei, scale OneHot for enterprise customers, hire for commercial and developer-relations roles and grow its Cambridge research team.
The push around decision models has accelerated since TypeSafe launched Jev in September 2026 and attracted attention from large cloud providers and AI vendors. TypeSafe itself raised substantially more capital in October 2026, underscoring divergent approaches within AI: large, heavily funded platforms versus lightweight models aimed at low-latency, structured decision-making.
Sqwish’s position — small teams, modest capital and open-weight releases — is a reminder that not every AI workload benefits from monolithic models. The company is pitching kettei as a practical middle layer that can lower compute costs and surface uncertainty estimates to downstream systems.
This funding round also highlights the role of university-backed funds and Innovate UK in seeding early-stage AI ventures in the Cambridge ecosystem. As decision models become a mainstream component of agent architectures, UK and European AI startups and investors will be watching whether lightweight, calibrated models can scale in production alongside larger generative systems.
| Investors | Investment Focus | Startup Investments | Round Size | Connect |
|---|---|---|---|---|
![]() | ||||
![]() Parkwalk( ) Parkwalk is a leading UK growth fund manager focused on investing in world-chang... London | ||||
![]() Vento Ventures( ) Turin, Italy | ||||
| All investors | All investor sectors | All funded startups | All funding rounds |
| Angel Investors | Location | Connect | Investment Focus | Startup Investments |
|---|---|---|---|---|
Steve Young is the Chairperson and an Emeritus Professor of Information Engineer... | ||||
Experienced business angel and exited founder, specialising in technology, AI, m... | ||||
| All angel investors | All locations | All investor sectors | All funded startups |
*These angel investors have publicly mentioned their interest in angel investing or are members of angel networks. (Last updated: November 2026)
Click here for a full list of 7,589+ startup investors in the UK