Every October, Gartner's IT Symposium/Xpo produces a ranked list of strategic technology trends, and every October that list ends up in a few thousand board decks. This year's edition, presented during the October 19–22 symposium and published on October 20, 2026, opened with an entry that will send a lot of people to a search engine: Large Quantitative Models.
The list
The ten trends Gartner named for 2027 are:
- Large Quantitative Models
- Speech Reasoning Models
- World Models
- AI Context Platforms
- Agent Coding Harness
- AI Agent Identity
- Autonomous Exposure Remediation
- Sovereign Tech
- Disposable Applications
- Wetware as a Service
Three of the top three are new kinds of model. That is a notable shift from lists of recent years, which were dominated by agents and platforms. The message is that the language-model era is giving way to a portfolio of specialised models, each built for a different kind of reality: numbers, speech and physical space.
The argument
Gartner's case for LQMs is short and pointed. General-purpose language models deployed on numerical tasks introduce severe accuracy and compliance risk, because what they produce is, in the end, a probabilistic guess. LQMs — described as modular, hybrid neural-network architectures for quantitative data — are presented as the corrective, moving decision-making from reactive troubleshooting to proactive simulation.
The language-model era is giving way to a portfolio of specialised models — one for words, one for speech, one for space, and one for numbers.
Two details in that framing matter. First, Gartner leads with risk, not capability. The pitch to a CIO is not "LQMs are cool", it is "you are already using LLMs for numerical work, and that is a liability". Second, "modular" and "hybrid" signal that Gartner does not expect one monolithic model to win. Expect the market to be described in terms of stacks and components rather than a single champion.
Why regulated industries
Gene Alvarez, Distinguished VP Analyst at Gartner, tied the trend directly to regulated sectors, arguing that the approach lets organisations safely deploy autonomous forecasting in strictly regulated industries such as finance and healthcare.
That is a meaningful choice of words. Regulators in finance and medicine care about reproducibility, auditability and quantified uncertainty. A language model that gives a slightly different answer each time is hard to defend in front of a model-risk committee. A quantitative model with a declared error bar, trained on traceable data, is much easier to validate — and model validation is exactly the profession that will decide how fast LQMs enter banks and hospitals.
How it fits the other nine
Read as a set, the list tells a coherent story, and LQMs sit at its centre:
- World Models are the spatial sibling of LQMs — physics-first systems that let agents simulate future states before acting. The boundary between "quantitative" and "world" models will blur fast.
- AI Context Platforms address the data problem. Gartner predicts that by 2027, organisations prioritising semantics in AI-ready data could raise agentic AI accuracy by up to 80% and cut costs by up to 60% (Digital Today). LQMs live or die by the same discipline.
- Sovereign Tech matters because quantitative models are trained on the crown jewels — lab data, trading data, patient data. Where that data lives will shape who can build them.
- Wetware as a Service, at number ten, is the far horizon: computation on biological substrates. It is a reminder that "compute" itself is diversifying.
What happens next
Trend lists do not create markets, but they do create budget lines. Based on how previous Gartner terms travelled, a reasonable expectation for the coming months is:
- Relabelling. Vendors with forecasting, simulation or scientific-ML products will adopt "LQM" in their positioning — some with justification, some without. Ask for benchmarks against measured ground truth.
- RFP language. "Quantitative model" and "LQM" will begin to appear in enterprise procurement, especially in finance, pharma and industrial engineering.
- Talent pull. Demand for people who can bridge scientific computing and modern ML will rise faster than supply.
- Scrutiny. The first high-profile LQM failure — a confident number that was wrong — will arrive, and it will be instructive.
The longer view is in The road to 2030.