Under the rising costs of LLMs, a new wave of ultra-fast AI decision models is disrupting enterprise AI with high performance matched by high efficiency and sub-31ms local inference. Built on Qwen3.5-2B, FLock.io's 1.9-billion-parameter decision model trained on synthetic data is for high-volume tasks that need quick decisions and data classification without compromising data sovereignty and privacy.

London, United Kingdom--(Newsfile Corp. - October 21, 2026) - FLock.io, a privacy-preserving AI solution provider, has launched an open-source small decision model built on Alibaba Cloud's Qwen3.5-2B. Called THIS/THAT Model, it is small enough to run on standard developer hardware and edge devices, such as smartphones and laptops, meaning data can stay local and latency can be ultra low.

To view an enhanced version of this graphic, please visit:
https://images.newsfilecorp.com/files/8871/318917_4797bba404ad6b19_001full.jpgFLock.io

The lightweight THIS/THAT model significantly undercuts operational costs and latency, despite being just 1.88B parameters. Three weeks after going live, it has already processed 2.5 million decisions. The launch positions FLock.io in a wave of decision models that have launched in recent weeks. In contrast with chatbots, these are designed for high-volume enterprise decision-making and classifications, where the slow, costly process of generating open-ended conversational text is unnecessary.

"For many critical enterprise tasks, generating open-ended text is an unnecessary cost and performance bottleneck," said Jiahao Sun, CEO of FLock.io. "With THIS / THAT Model, we are demonstrating that specialised 2B-parameter models running locally - all while keeping data strictly on-device."

THIS/THAT can be deployed locally for full data privacy and sovereignty

Instead of sending data to central cloud servers, FLock.io's open-source architecture enables true edge deployment. Running locally on developer hardware, laptops, or enterprise edge nodes allows organisations in regulated sectors - such as finance, healthcare, and defense - to execute real-time automated workflows completely offline, guaranteeing data sovereignty and regulatory compliance.

Edge computing is a distributed computing model that processes and stores data physically close to where it is created - such as on an IoT device, local gateway or edge server - rather than sending all raw data to a centralised cloud or distant data center.

For regulated sectors such as governments, banking institutions, and healthcare providers, it's essential to keep data secure and comply with regulation. By executing decisions locally, THIS / THAT Model ensures sensitive data never leaves the user's machine. It requires no central GPU servers and no cloud connectivity.

High performance at the edge

In third-party evaluations, version 1.2 achieved 94.1% accuracy on 68 local-geometry questions, and a 87.8% success rate across 1,710 complex decision-making tasks on laptop-based benchmarks. This was a rapid improvement from 1.0's tests just days before, achieved through optimising training data rather than increasing model size.

THIS / THAT Model's specs

THIS / THAT Model was built on Qwen3.5-2B: a tiny, edge-friendly foundation model that can be run locally on smartphones and laptops. The local inference latency is approximately 31.4 milliseconds. It is lightweight, with only 1.88B parameters and zero generated output tokens. It was trained on Alibaba Cloud using synthetic training data generated using Qwen Max. It drew on key findings from FLock.io's research papers, GraphSynth and CircuitSynth, accepted into the prestigious ACL 2026 conference.

Background on Qwen and FLock.io's partnership

The launch deepens the relationship that began between Qwen, a tier 1 LLM developed by Alibaba Cloud, and FLock.io in April 2025. Summer last year marked the first time a major Web2 centralised LLM has chosen to deeply integrate with a decentralised training platform.

About FLock.io

FLock.io is an AI research and infrastructure company headquartered in London. It pioneers enterprise-grade federated learning and distributed AI solutions. Its decentralised FL architecture and production-ready platforms (AI Arena, FL Alliance, FOMO and the API Platform) enable organisations to train and deploy their own custom AI models on local hardware while maintaining full data privacy, model ownership, and regulatory alignment by design.

To view the source version of this press release, please visit https://www.newsfilecorp.com/release/318917