Osinto - An Independent Intelligence House
IndependentIntelligence
Osinto is an independent intelligence house. We conduct investigations, collect commercial intelligence and evaluate AI models.
Intelligence
you can trust
Intelligence you can't trust is worse than useless. That's why Osinto's approach is grounded in provenance and defensibility - whether we're scoring the credibility of a source or the reliability of an AI model.
Investigations & Intelligence
We provide clients with clarity - on counterparties and countries, capital flows and markets.
- Investigations & Due Diligence
- Geopolitical & Market Entry Advisory
- Market & Capital Intelligence
AI Evaluations
We bring investigative discipline to machine learning model evaluation and hands-on experience to enterprise AI.
- AI Evaluations & Benchmarking
- Applied AI
We build tools in-house to encode auditable discipline in our work. Cloud Empire™ is our entity mapping platform. Our AI Evaluations are built on the Inspect framework.
Commissioned by frontier operators and capital allocators
AI Model Selection
The CIDER Press (Can It Do Expert Research?) is Osinto's first published benchmark; evaluating the performance of open weights models on Apple Silicon. On a single, air-gapped MacBook Pro we test leading Chinese, European and North American Large Language Models (LLMs) using a lightweight, tool-calling agentic harness - against a held-out set of real-world knowledge work tasks.
Dispatches from the frontier
Now a podcast Apple Podcasts Spotify Amazon Music Audible
