I was reading about vertus ai, they are challenging llm based technology with dynamic reasoning neural topology like system. Can you check what they actually do and simplify their role and offerings
Searching the Web
Vertus AI dynamic reasoning neural topology LLM alternative
Searching the Web
Vertus AI company offerings products 2025 2026
Reading Web Page
Vertus AI technology architecture offerings products intelligence API
Here's a clear breakdown of what Vertus AI actually is and does:
What is Vertus AI?
Vertus is a frontier AI company founded by Julius Franck, Alex Foster, and Michal Prywata. Their core bet: build AI that genuinely reasons, rather than predicts. They position themselves against standard LLMs (like GPT or Claude), which they argue are essentially very sophisticated pattern-matching systems — not true reasoners.
The Core Tech: Dynamic Neural Topology
The term you read about is their signature concept. Here's what it means in plain terms:
| Traditional LLMs | Vertus "Superintelligence" | |
|---|---|---|
| Architecture | Fixed — same neural pathways used for every query | Dynamic — forms new neural connections per query |
| What it does | Retrieves patterns from training data | Builds a custom reasoning pathway on-the-fly for each problem |
| Analogy | A giant lookup table | A brain forming new synapses for each thought |
Simple question → lightweight pathway. Complex problem → dense, interconnected topology. The idea mirrors how biological brains allocate different cognitive resources depending on task complexity.
Their models are measured in "Meganeurons" (millions of interconnected nodes), with current public releases at 8Mn and 16Mn architectures, and more advanced ones reportedly in development.
What Do They Actually Sell?
1. AI Trading & Financial Intelligence (their flagship)
- Their systems manage $600M–$1B+ in daily institutional trading volume
- Claimed 51.15% return in 2025 with a 2.13 Sharpe ratio (independently audited)
- Uses proprietary broker money flow data invisible to public markets
- Serves hedge funds, family offices, and asset managers
- They use financial markets as a "proving ground" — the logic being that if AI can make correct decisions in milliseconds with real money on the line, it's genuinely reasoning, not hallucinating
2. Intelligence API
- Developers can build Vertus's reasoning architecture into their own apps
- Each user gets their own learning AI instance (multi-tenant design)
- Targeted at: code generation, data analysis, business strategy, research, writing
3. General-Purpose Reasoning Models (public release)
- 8Mn-General — analysis, research, strategy, problem-solving
- 8Mn-Coding — software development and debugging
- 16Mn-General — deeper multi-domain reasoning
- 16Mn-Coding — advanced system design
4. Research Partnerships
- Collaboration on consciousness research and ethical AI
The Honest Picture
Vertus's marketing language is bold ("world's first nature-inspired artificial minds," "AI that actually thinks"). A few things to keep in mind:
- Their trading performance claims are audited by a CPA firm, which adds some credibility — but "dynamic neural topology" as they describe it isn't a widely peer-reviewed architecture, so independent validation of the underlying technology is limited.
- They were initially developed and tested in the Isle of Man, a jurisdiction known for lighter regulatory oversight.
- The broader "expand beyond finance" narrative is aspirational — the financial product is real and operational; the general AI platform is a newer, public-facing extension of that core work.
In short: Vertus is a finance-first AI company that built proprietary reasoning systems for institutional trading, is now productizing that tech as a general API, and uses "dynamic neural topology" as a marketing label for their adaptive, query-specific inference architecture.
This is a shared conversation. Sign in to Orris to start your own chat.