Multiverse Computing Explained: AI Technology, Products & Use Cases

Multiverse Computing is a European AI and quantum-software company focused on making artificial intelligence smaller, more efficient, secure, and easier to deploy. Its technology combines quantum-inspired methods, tensor networks, model compression, and AI optimization to reduce the computing resources required to run advanced models.

The company’s best-known AI technology is CompactifAI, a model-compression platform designed to shrink large language models and other AI systems while retaining much of their original capability. Multiverse Computing also develops Singularity, a platform for AI and optimization, and has expanded into compressed AI models, sovereign AI infrastructure, edge AI, and enterprise AI governance.

More recently, Multiverse Computing has introduced its own AI models, including Quasar 438B and Quasar 1.1 438B, while continuing to make compressed versions of models from companies such as Meta, Mistral, NVIDIA, Qwen, Microsoft and Z.ai available through its ecosystem.

Quick Summary –

• Multiverse Computing is a European AI and quantum-software company focused on efficient AI.
• Its CompactifAI technology compresses AI models while aiming to reduce inference costs and improve speed.
• Singularity provides quantum and quantum-inspired optimization tools for enterprise applications.
• The company is developing Quasar AI models for reasoning, coding, and enterprise AI agents.
• Its technologies support edge AI, private AI, sovereign AI, and efficient model deployment.
• Multiverse Computing serves industries including finance, healthcare, manufacturing, energy, and cybersecurity.
• Overall, the company is focused on making large AI models smaller, faster, and more efficient.

What Is Multiverse Computing?

Multiverse Computing is a technology company founded in 2019 and headquartered in Donostia-San Sebastián, Spain. It originally focused on quantum software and optimization before expanding its work into AI model compression and efficient AI deployment.

The company’s central idea is straightforward: increasingly capable AI models require enormous amounts of computing power, memory, energy and infrastructure. Instead of assuming that every AI workload needs increasingly large hardware clusters, Multiverse Computing develops technologies intended to make models more compact and efficient.

Its current business can broadly be understood through several areas:

• AI model compression
• Efficient AI inference
• Edge and on-device AI
• Sovereign and private AI deployment
• Quantum-inspired AI
• AI and mathematical optimization
• Enterprise AI governance
• AI model development

The company says its solutions are used across industries including finance, energy, manufacturing, healthcare, aerospace, cybersecurity, defense and engineering.

Why Does Multiverse Computing Matter for AI?

Large AI models can deliver strong performance, but deploying them can require expensive GPUs, large memory footprints, high energy consumption and substantial data-center infrastructure.

That creates a practical problem for organizations that want to run AI:

How can a company use advanced AI without continually increasing its infrastructure requirements?

Multiverse Computing approaches the problem from the model-efficiency side.

Its CompactifAI technology is designed to compress model weights and reduce the size of AI models. The company’s current website says its compressed models can deliver 50–80% lower inference costs, up to 2× faster inference, and close to 100% accuracy retention, although actual results vary by model and workload.

This matters particularly when AI has to operate outside a giant cloud data center.

Examples include:

• AI running on enterprise servers
• Private AI deployments
• On-premise systems
• Mobile devices
• Autonomous systems
• Industrial equipment
• Telecom infrastructure
• Satellites
• Vehicles
• Edge computing environments

The goal is not simply to make AI models smaller. It is to make advanced AI more deployable.

What Is CompactifAI?

CompactifAI is Multiverse Computing’s proprietary AI model-compression technology.

In simple terms, it takes an AI model and applies mathematical techniques designed to reduce its computational and memory requirements while attempting to preserve useful capabilities.

Multiverse Computing says CompactifAI uses quantum-inspired tensor networks to compress foundation models, including large language models.

The company describes several benefits:

• Reduced model size
• Fewer parameters
• Lower memory requirements
• Faster inference
• Lower computational costs
• Lower energy consumption
• Easier deployment on private infrastructure
• Greater portability across hardware

The technology is particularly relevant because conventional model compression can involve trade-offs between model size and performance.

Multiverse Computing has reported examples in which its compressed models retain high accuracy while substantially reducing their size and resource requirements. For example, the company reported compressing Llama 3.1 8B and Llama 3.3 70B by 80%, with 40% faster inference and a reported 50% cost reduction in those specific evaluations.

Those figures should be treated as company-reported results for the specified models and testing conditions, rather than universal performance guarantees for every CompactifAI deployment.

How Does CompactifAI Work?

At a high level, CompactifAI uses mathematical representations to identify ways of representing model information more efficiently.

One of the technologies behind the approach is tensor networks.

A neural network contains a very large number of numerical parameters. Storing and processing all of them in their original form can require significant memory and compute.

Tensor-based representations can restructure this information into more compact mathematical forms.

A simplified workflow looks like this:

Original AI model → mathematical decomposition → compression/optimization → evaluation → optimized model → deployment

The objective is to reduce the model’s computational footprint without making it unusable for its intended task.

This is different from simply deleting random parameters.

The compression process has to consider whether important capabilities survive the transformation. Multiverse Computing describes CompactifAI as a broader optimization pipeline rather than merely a file-size reduction technique. Its newer Quasar work, for example, combines compression with additional training, tuning and optimization.

What Products Does Multiverse Computing Offer?

Multiverse Computing’s product ecosystem has expanded considerably beyond its original quantum-computing focus.

The major products and technologies include CompactifAI, Singularity, Quasar models, CompactifAI API, CompactifAI App, Multiverse Computing Foundry, and SentinelAI.

The company is therefore moving from being primarily known for quantum-inspired optimization and model compression toward a broader efficient and sovereign AI stack.

What Is Singularity?

Singularity is Multiverse Computing’s software platform for AI, machine learning and optimization.

Its technology combines quantum and quantum-inspired algorithms to address complex optimization and industrial AI problems. The platform includes high- and low-level APIs and is designed for deployment within industrial workflows.

Singularity has been applied to areas such as:

• Finance
• Manufacturing
• Energy
• Healthcare
• Cybersecurity
• Defense
• Machine vision
• Predictive maintenance
• Fraud detection
• Portfolio optimization
• Anomaly detection

For example, Multiverse Computing describes applications involving manufacturing equipment, financial anomaly detection, healthcare monitoring and autonomous-vehicle machine vision.

This is where the company’s original quantum-computing expertise remains relevant. Singularity is not simply an LLM product; it addresses broader mathematical optimization and machine-learning problems.

What Is the CompactifAI API?

The CompactifAI API provides developers with access to a catalog of AI models through an API rather than requiring them to manage the underlying infrastructure themselves.

The current platform includes models from Multiverse Computing and selected third-party model providers, with an emphasis on efficient inference and coding workloads.

Multiverse Computing says developers can use the API for:

• Model experimentation
• Application development
• Coding agents
• AI assistants
• Production inference
• Enterprise deployments

The platform supports usage-based billing and offers private deployment options for organizations that need more control over infrastructure and data.

Multiverse has also added models from other AI developers. For example, it announced support for NVIDIA’s Nemotron 3 family through the CompactifAI API.

What AI Models Does Multiverse Computing Develop?

Multiverse Computing has increasingly moved from compressing other organizations’ models to developing and publishing its own models.

One of the most prominent examples is Quasar 438B, introduced in September 2026.

Quasar 438B

Multiverse Computing describes Quasar 438B as a flagship reasoning model designed for enterprise-scale agents and coding.

The company says it scores highly on several Artificial Analysis evaluations and reported a score of 43 on the Artificial Analysis Intelligence Index v4.1.1. It also highlighted its performance on terminal-based coding tasks.

Quasar 438B is particularly interesting because Multiverse is applying its efficiency strategy to a model in the 400-billion-plus parameter class.

Rather than treating model size alone as the objective, the company is attempting to combine large-model capabilities with more efficient deployment.

Quasar 1.1 438B

On September 15, 2026, Multiverse Computing introduced Quasar 1.1 438B, a rebuilt version of its flagship coding model.

The company says the model was built from GLM-5.2, an open-weights model from Z.ai, and transformed through the CompactifAI pipeline. The updated version adds broader retraining, output-length tuning and quantum-generated data to the development process.

Multiverse says part of the new training data was generated using a hybrid quantum language model running circuits on an IBM Quantum System Two with a 156-qubit IBM Heron processor.

This is an important distinction: quantum computing is being used as part of the model-development pipeline, rather than claiming that Quasar itself is simply a quantum computer-based LLM.

What Is the CompactifAI App?

The CompactifAI App brings compressed AI models to mobile and edge devices.

Multiverse Computing launched the app in March 2026 with the goal of allowing users to run advanced AI models locally and offline.

The app can:

• Run supported AI models locally
• Operate without an internet connection
• Keep sensitive queries on the device
• Route more demanding tasks to cloud-based models
• Support environments with unreliable connectivity

This makes the technology particularly relevant to field workers and organizations operating in environments where cloud connectivity is limited or data cannot easily leave the organization’s infrastructure.

Potential applications include:

• Healthcare
• Legal work
• Defense
• Manufacturing
• Field operations
• Enterprise mobile applications

The broader concept is AI at the edge: instead of sending every request to a remote data center, some AI processing happens directly on the device.

What Is SentinelAI?

Multiverse Computing is also expanding into AI governance through SentinelAI.

Announced in July 2026, SentinelAI is designed as an enterprise control layer positioned between AI applications, users, agents and the models they use.

Its intended functions include:

• Detecting sensitive information
• Identifying credentials or confidential documents
• Detecting prompt-injection and jailbreak attempts
• Enforcing organizational AI policies
• Monitoring AI usage
• Recording activity for audits
• Controlling what information is sent to models

A major feature of SentinelAI is that it is designed to be model-agnostic.

That means an organization can potentially apply the same governance layer to commercial models, open-source models, internal models and Multiverse’s own models.

The system is designed for on-premise deployment, allowing governance controls and audit records to remain within the organization’s environment.

What Is Multiverse Computing Foundry?

Multiverse Computing is also developing a broader Foundry platform for sovereign AI infrastructure.

The company describes Foundry as a unified platform intended to bring together:

• AI model discovery
• Model deployment
• GPU infrastructure
• Workload monitoring
• Cost monitoring
• Serverless AI services

The platform is listed as coming soon on the company’s current website.

The idea is significant because Multiverse Computing is increasingly positioning itself beyond model compression.

Its broader proposition is becoming:

Efficient models + infrastructure + deployment + governance + sovereignty

rather than simply:

Smaller AI models.

What Are the Main Use Cases for Multiverse Computing?

Multiverse Computing’s technologies can be applied across a wide range of industries.

1. Enterprise AI

Companies can use compressed models to reduce the infrastructure required for AI inference.

This can be useful when organizations want to extend existing hardware instead of immediately buying additional GPUs.

Multiverse says its enterprise solutions can reduce GPU requirements, energy consumption and infrastructure costs while allowing organizations to deploy models within their own environments.

2. Edge AI

Edge AI is one of the most natural applications for model compression.

Devices such as:

• Cameras
• Vehicles
• Industrial machines
• Smartphones
• Satellites
• Drones
• Telecom equipment

have significantly fewer computing resources than large AI data centers.

A smaller model can make local inference more practical.

3. Private and Sovereign AI

Organizations handling sensitive information may not want to send all AI workloads to public cloud services.

Compressed models can make it easier to run AI:

• On-premise
• Inside a private cloud
• At the edge
• In air-gapped environments

Multiverse Computing’s recent partnership activity has increasingly focused on this area, including deployments for regulated industries.

4. Coding Agents

Multiverse has positioned its CompactifAI API for coding-agent workloads.

In May 2026, the company announced integrations allowing developers to use CompactifAI as a backend for leading coding agents without changing their existing IDE or workflow. Multiverse reported that its internal evaluations showed up to 75% lower cost per token in the tested setup.

Again, that figure is an internal company evaluation rather than a universal price reduction for every coding-agent configuration.

5. Manufacturing

Manufacturing environments generate large quantities of sensor and machine data.

AI can be used for:

• Predictive maintenance
• Defect detection
• Machine monitoring
• Anomaly detection
• Production optimization

Multiverse Computing’s Singularity platform has been used for several of these industrial applications.

6. Finance

Financial institutions can use AI and optimization technologies for:

• Fraud detection
• Risk analysis
• Portfolio optimization
• Trading systems
• Anomaly detection

These workloads can benefit from both efficient AI and specialized mathematical optimization.

7. Healthcare

Multiverse Computing has described AI applications involving intensive-care monitoring and predictive systems.

The attraction is straightforward: healthcare environments can generate enormous volumes of sensor data, while latency, privacy and reliability can be important considerations.

8. Cybersecurity

Multiverse’s AI technologies can also be used for anomaly detection and threat intelligence.

The objective is to identify unusual patterns in large amounts of network or system data that may indicate malicious activity.

Multiverse Computing vs Traditional AI Model Compression

Model compression is not unique to Multiverse Computing.

AI developers already use techniques such as:

• Quantization
• Pruning
• Knowledge distillation
• Low-rank methods
• Weight sharing
• Architecture optimization

The difference is that Multiverse Computing’s approach centers heavily on tensor networks and quantum-inspired mathematical techniques.

These techniques are not necessarily mutually exclusive. Modern AI optimization pipelines can combine multiple approaches.

What Makes Multiverse Computing Different?

The company’s differentiation comes from combining several areas that are usually treated separately.

a) Quantum-inspired mathematics

Multiverse Computing has a background in quantum software and applies quantum-inspired mathematical approaches to AI optimization.

b) Model compression

CompactifAI focuses on reducing the computational footprint of AI models.

c) Edge deployment

Smaller models can potentially run in environments where conventional large models are impractical.

d) Sovereign AI

The company increasingly emphasizes local infrastructure, data control and European AI sovereignty.

e) AI governance

SentinelAI extends the company’s stack into enterprise monitoring, policy enforcement and security.

f) Proprietary models

With Quasar, Multiverse Computing is moving further into developing its own AI models rather than only optimizing models created elsewhere.

Taken together, these areas explain why the company is increasingly describing its strategy around efficient and sovereign AI.

Is Multiverse Computing a Quantum Computing Company or an AI Company?

The answer is both, but its current business is increasingly centered on efficient AI.

Multiverse Computing was founded around quantum software and optimization. Its Singularity platform continues to use quantum and quantum-inspired algorithms.

However, the company’s recent growth has been strongly connected to AI model compression, efficient inference, AI models and sovereign deployment.

Its 2025 Series B and 2026 Series C fundraising announcements illustrate this transition. In July 2026, Multiverse announced a potential $570 million Series C aimed at scaling efficient AI from edge devices to data centers.

The company’s current website similarly places efficient AI at the center of its positioning while retaining Singularity and quantum software as important parts of its technology portfolio.

How Much Funding Has Multiverse Computing Raised?

Multiverse Computing announced a €189 million ($215 million) Series B in June 2025. The company said the round would support the development and commercialization of its AI compression technology.

In July 2026, it announced a $570 million (€500 million) Series C fundraising round at a reported $1.7 billion pre-money valuation. The company said the round could bring total funding to approximately $800 million, including previous rounds.

The Series C announcement also described deployments and partnerships spanning manufacturing, finance, energy, aerospace, cybersecurity, defense and healthcare.

These funding figures describe announced fundraising and should not be interpreted as revenue or profit.

What Are the Limitations of Multiverse Computing?

Multiverse Computing’s technology is promising, but several practical considerations matter.

1. Compression does not automatically preserve everything

Reducing a model’s footprint can affect capabilities depending on the compression method, model architecture and target workload.

Performance therefore needs to be measured on the specific application.

2. Company-reported benchmarks need context

Many performance and cost figures published by Multiverse Computing come from the company’s own testing.

Independent evaluation remains important when selecting a model or deployment strategy.

3. Not every AI workload needs compression

For some applications, using a hosted frontier model may be simpler than deploying and maintaining a compressed model.

Compression becomes particularly attractive when infrastructure cost, privacy, latency, hardware constraints or sovereignty are important.

4. Edge AI has hardware limitations

A compressed model can reduce requirements, but it does not make hardware unlimited.

Performance still depends on the device, memory, processor, accelerator, model architecture and workload.

5. Enterprise deployment requires more than a model

Organizations running AI locally also need:

• Monitoring
• Security
• Model updates
• Access controls
• Governance
• Infrastructure management
• Evaluation
• Compliance processes

This is one reason Multiverse Computing has expanded its offering beyond compression into deployment and governance.

What Is the Future of Multiverse Computing?

Multiverse Computing’s recent direction suggests that its strategy is moving toward a broader AI efficiency stack.

Several developments point in that direction:

• Larger compressed AI models
• Proprietary models such as Quasar
• On-device AI
• Sovereign AI infrastructure
• AI governance
• Enterprise deployment
• Partnerships with hardware companies
• Integration with coding agents
• Quantum-generated data and quantum-inspired model optimization

The company’s July 2026 Series C announcement specifically described plans to expand its model library, continue algorithm research and invest in sovereign AI infrastructure.

Its September 2026 Quasar 1.1 release also demonstrates how the company is experimenting with combining classical AI model development, CompactifAI compression and quantum-generated data.

That makes Multiverse Computing’s long-term story broader than simply making LLMs smaller.

The company is attempting to address a larger question:

How can organizations run increasingly capable AI while using less infrastructure, retaining more control over their data, and deploying models closer to where the data is generated?

Conclusion

Multiverse Computing is building technology around one of the central challenges facing modern AI: how to make increasingly capable models more efficient and deployable.

Its CompactifAI technology is the foundation of that strategy, using quantum-inspired tensor-network techniques to compress AI models and reduce their computational footprint. Singularity extends the company’s work into AI and optimization, while the CompactifAI API and App bring efficient models to cloud, private and edge environments.

The company’s newer Quasar models show another stage of its development: Multiverse Computing is no longer focused only on optimizing other organizations’ models but is also building its own AI systems.

Meanwhile, SentinelAI and the planned Foundry platform point toward a broader enterprise proposition involving governance, deployment, infrastructure and sovereignty.

Multiverse Computing is therefore best understood not simply as a quantum-computing startup or an AI model-compression company, but as a technology company working across efficient AI, model optimization, quantum-inspired computing, edge deployment and sovereign enterprise AI.

Frequently Asked Questions About Multiverse Computing

1. What is Multiverse Computing?

Multiverse Computing is a European technology company founded in 2019 that develops quantum, quantum-inspired and AI technologies. Its current focus includes AI model compression, efficient inference, edge AI, sovereign AI and enterprise AI solutions.

2. What is CompactifAI?

CompactifAI is Multiverse Computing’s AI model compression and optimization technology. It uses quantum-inspired techniques based on tensor networks to reduce model size and computational requirements while attempting to preserve useful model performance.

3. What is Quasar 438B?

Quasar 438B is a Multiverse Computing reasoning model designed for enterprise-scale agents and coding. The company introduced it in September 2026 and made it available through the CompactifAI API.

4. Does Multiverse Computing use quantum computing for AI?

Multiverse Computing uses quantum-inspired techniques in its AI technologies, particularly its model-compression work. Its September 2026 Quasar 1.1 release also incorporated data generated by a hybrid quantum language model running on an IBM quantum processor.

5. Can Multiverse Computing AI run offline?

Yes. The CompactifAI App is designed to run supported AI models locally on compatible devices without an internet connection. The company specifically positions this capability for privacy-sensitive, low-connectivity and field environments.

6. What industries use Multiverse Computing technology?

Multiverse Computing lists applications across finance, energy, manufacturing, healthcare, aerospace, cybersecurity, defense and other industrial sectors. Its technologies include optimization, anomaly detection, machine vision, AI model compression and edge deployment.

Also Read –

Quasar 1.1 438B Uses Quantum-Generated Data

AI Model Compression: Techniques, Benefits, Challenges & Use Cases