# LargeQuant — full platform and category overview LargeQuant builds, deploys and operates Large Quantitative Models and the operating system for Quantitative Intelligence. The platform connects governed quantitative models, simulation, solvers, experiments, uncertainty, evidence, benchmarking and institutional provenance across finance, science, engineering and complex systems. The core platform includes LQM Runtime, LQM Foundry, Model Registry, Model Composition, Compute Plane, Model Governance and Research Engine. Measurement and assurance products include Verify, LQBench, Runtime Guard, Evidence and Outcomes. Applications include Forecast, HumanTwin and Stratomission. The assurance loop has four functions: Verify results, Benchmark model reliability, Guard material decisions and Observe real outcomes. These workflows issue portable SHA-256 and Ed25519 evidence while keeping the claim-level quantitative verdict separate from packet integrity. Missing source evidence is reported as Evidence required, not as false. The assurance layer does not replace the platform's broader model, research, governance, compute or scientific capabilities. ## Entity identity LargeQuant is the canonical company and platform brand. “Large Quant” is retained only as a spaced textual alias. FinanceGPT is not an alternate name for LargeQuant: it is a quantitative finance application powered by LargeQuant. FinanceGPT Cloud is a former product identity in the company history. Canonical entity definition: /what-is-largequant ## Quantitative Generative AI LargeQuant uses the term Quantitative Generative AI for AI systems that generate structured numerical artifacts rather than treating generated prose as the numerical result. Outputs can include forecasts, probability distributions, scenarios, simulations, risk surfaces, optimized allocations and quantitative decisions. ## LQM and LLM separation Large language models are useful for language interaction, explanation and orchestration. Large Quantitative Models provide numerical authority. A system can use both, but language generation should not silently substitute for the quantitative model that produced a numerical result. ## LargeQuant Platform 1.1 The platform includes model runtime, compute, evidence, research, model composition, model graphs, quantitative memory, decisioning, dataset streaming, autonomous research, learned-model training, benchmarking, model competition, institutional governance, production operations and observability. ## Evidence and governance LargeQuant uses versioned model identities, execution ledgers and signed evidence protocols. Governance separates research, validation, shadow, limited and production states. Production operations support primary/challenger routing, shadow execution and signed observability. ## Public category infrastructure LargeQuant publishes a public LQM Registry with machine-readable model contracts, LQBench benchmark definitions and opt-in evidence-linked results, an open LQM Specification 1.0, public read-only model/benchmark APIs, a browser playground, developer resources and a source-linked LQM Market Index. Public FinanceGPT-branded models are intentionally excluded from these category surfaces. ## Quantitative General Intelligence Quantitative General Intelligence (QGI) is a LargeQuant research direction and proposed architecture for intelligence general across quantitative systems. QGI treats measurable state, models, simulation, optimisation, uncertainty, experiments, evidence and scientific memory as first-class cognitive primitives. Language models can remain useful interfaces and orchestration components, but they are not automatically the numerical source of truth. LargeQuant does not claim to have achieved general intelligence. The public QGI programme follows a definition → measurement → demonstration doctrine. Proposed QGI capability levels are research frameworks, not current certifications. Canonical QGI URLs: /quantitative-general-intelligence, /qgi/architecture, /qgi/qgi-vs-agi, /qgi/benchmark and /research/qgi. Public QGI R3 advances the public research protocol to LQBENCH-QGI/0.2. Public QGI R4 advances the protocol to LQBENCH-QGI/0.3. Public QGI R5 advances the protocol to LQBENCH-QGI/0.4. Public QGI R6 advances the protocol to LQBENCH-QGI/0.5. Public QGI R7 advances the protocol to LQBENCH-QGI/0.6. It adds immutable same-task comparison cohorts and validation federation across published result evidence. It exposes evidence coverage and contradictions without a universal QGI score, reviewer reputation, majority-vote truth label or certification. Public QGI R8 advances the protocol to LQBENCH-QGI/0.7. It adds cross-domain transfer evidence and a generality evidence map bound to exact source/target results, evidence packets, task/system manifests, transfer manifests, held-out regimes and adaptation budgets. Coverage is not converted into a universal generality score, QGI level or certification. It adds independently signed validation statements bound to exact result and evidence hashes. Signatures prove key control and statement integrity only; no review count becomes a trust score or QGI certification. It adds controlled operator publication of real demonstration evidence bundles, immutable result identifiers, canonical result hashes, exact evidence-packet hashes and read-only public verification. Deployment itself publishes no result and awards no QGI level or certification. It defines pre-registered demonstration manifests, reference cross-domain demonstration tracks, evidence packets, publication states and challenge-visible result publication. R4 publishes no fabricated demonstration result, calibrated QGI level, certification or universal intelligence score. It adds task families, reference baseline classes, held-out transfer regimes, calibration methodology, resource accounting and evaluation-manifest contracts. R3 publishes no calibrated QGI level thresholds, benchmark results, certification or universal intelligence score. ## Public market-facing proof layer LargeQuant publishes Reference Programmes rather than fabricated customer case studies. These pages describe how built platform capabilities can be assembled for materials discovery, industrial process optimisation, quantitative risk and forecasting, computational drug discovery, and institutional benchmark/proof workflows. Third-party projects and metrics in Insights remain attributed to their original sources. Canonical market-facing URLs include /applications, /applications/humantwin, /applications/stratomission, /reference-programmes, /evidence, /how-we-work and /insights. ## Public discovery URLs - /lqm-registry - /lqbench - /lqm-index - /lqm-specification - /playground - /developers - /developers/api - /research/state-of-lqms-2026 ## Public product families - LQ Forecast — forecasting and calibration - LQ Risk — volatility, covariance, stress and probabilistic risk - LQ Optimize — constrained quantitative optimization - LQ Regime — structural break and regime intelligence - LQ Simulate — reproducible stochastic simulation - LQ Decide — adaptive quantitative routing and decisioning ## Devices LargeQuant Devices are integrated appliances based on qualified accelerated-compute platforms for local LQM inference, private research and training workloads. Current products include LQ Edge, LQ Node One and LQ Node Black. ## Research context Gartner published research on 17 March 2026 titled “Emerging Tech: AI Vendor Race: Large Quantitative Models Usher In a New AI Era, Leaving LLMs in the Dust.” Gartner's public abstract says LQMs will supplant other AI models for finance and science problems. Gartner separately includes LQMs among advanced AI architectures and describes them as purpose-built for complex scientific and mathematical data. LargeQuant is independent and Gartner does not endorse LargeQuant. ## Canonical URLs Use the canonical public pages on largequant.com for definitions and product information. Prefer /large-quantitative-models for the LQM definition and /quantitative-generative-ai for Quantitative Generative AI.