Loading organizations...
Trismik offers QuickCompare, an adaptive evaluation platform for enterprise Large Language Models (LLMs). The platform guides organizations in selecting optimal AI models by providing transparent trade-offs across accuracy, latency, and cost. It processes real evaluation data, offering actionable insights to ensure robust model performance within production environments.
Founded in 2023 as a University of Cambridge spinout, Trismik was established by Rebekka Mikkola, Professor Nigel Collier, and Marco Basaldella. Mikkola, an experienced founder, and Professor Collier, a renowned NLP expert, identified conventional LLM evaluation as inefficient at uncovering real-world risks.
Trismik supports teams deploying and integrating AI models, enabling data-driven selection. Its immediate goal is to simplify model choice through detailed performance understanding. The company’s long-term vision involves constructing intelligent decision layers for AI systems, fostering automated, scalable evaluation and continuous enhancement of AI capabilities.
Trismik has raised $3.0M across 1 funding round.
Trismik has raised $3.0M in total across 1 funding round.
Trismik has raised $3.0M across 1 funding round. Most recently, it raised $3.0M Seed in September 2025.
| Date | Round | Lead Investors | Other Investors | Status |
|---|---|---|---|---|
| Sep 1, 2025 | $3M Seed | John Spindler | Antler, Fuel Ventures, Christine Martin, Fund F, Parkwalk Advisors, Vento Ventures, Ventures Together | Announced |
Trismik is a science-driven AI startup developing an adaptive evaluation platform for large language models (LLMs), applying psychometric principles like Item Response Theory (IRT) and Computerized Adaptive Testing (CAT) to assess AI performance on factuality, bias, toxicity, reasoning, and safety[1][2][3][4]. Spun out of the University of Cambridge, it serves AI developers, researchers, and enterprises by solving the bottleneck of slow, inconsistent traditional LLM testing, delivering results up to 180x faster at lower cost while maintaining high precision[2][3]. Recently raising £2.2M in pre-seed funding in September 2025, Trismik shows strong early momentum with platform pilots, enterprise partners in media, telecom, and tech, and plans for a full enterprise release in 2026[1][2][3].
Trismik was co-founded in early 2025 by CEO Rebekka Mikkola, Chief Scientist Professor Nigel Collier, and CTO Marco Basaldella, emerging as a spinout from the University of Cambridge's Language Technology Lab[1][2]. Collier and Basaldella's prior collaboration at Cambridge, including Basaldella's postdoctoral work and his later experience at Amazon on foundation models, provided technical depth, while Mikkola led the venture[2]. The idea stemmed from the growing need for scalable LLM evaluation amid benchmark saturation and rising demands for trust in AI, with the team joining Cambridge's START 2.0 accelerator in early 2025 and securing a licence from Cambridge Enterprise by May[2][3]. Pivotal early traction included launching the first platform version, onboarding enterprise partners, and closing the £2.2M pre-seed round led by Twin Path Ventures[1][2].
Trismik rides the explosive growth of LLMs powering products across industries, where evaluation has become a critical bottleneck due to benchmark saturation in reasoning, math, coding, and general knowledge[3]. Timing is ideal amid surging AI adoption pressures for safety, reliability, and trust, with market forces like regulatory demands for compliance and the need for faster dev cycles favoring adaptive tools that cut costs and scale[1][2][3]. By setting a new gold standard in psychometric-inspired testing, Trismik influences the ecosystem, enabling Cambridge's AI contributions globally and empowering builders to iterate confidently in high-stakes environments like media and telecom[2][3].
Trismik is poised to redefine AI evaluation standards, with near-term focus on enterprise onboarding by late 2025, full platform launch in early 2026, and expansions into advanced experimentation features[1][3]. Trends like AI safety regulations, multimodal models, and trust-at-scale demands will propel its growth, potentially evolving it into an end-to-end LLM ops hub[1][3][4]. As benchmark limitations intensify, Trismik's science-grade precision could cement its influence, transforming slow testing into a competitive edge for AI teams worldwide—bridging Cambridge innovation to practical, trustworthy AI deployment[2][3].
Trismik has raised $3.0M in total across 1 funding round.
Trismik's investors include John Spindler, Antler, Fuel Ventures, Christine Martin, Fund F, Parkwalk Advisors, Vento Ventures, Ventures Together.
