Featured analysis
What Is AI Governance?
A working definition of AI governance, how it differs from safety, ethics and compliance, and what separates a governance policy from a governance control.
Read full analysisEvidence. Context. Clarity.
AI, explained with the evidence attached.
Independent analysis and structured reference data for understanding AI systems, models and the claims made about them. Every factual claim carries a source, and every unverified field says so.
Explore key topics
AI Governance
How organisations and regulators decide who is accountable for an AI system, what it may do, and how that is evidenced.
→Safety & Alignment
Techniques and operational practices for reducing AI risk, checking intended behaviour, and retaining meaningful human oversight.
→Systems & Methods
Architectures, training approaches and operational methods that shape how contemporary AI systems are built and used.
→Model Evaluation
Methods and benchmarks for measuring AI capabilities, reliability, limitations and the context in which results apply.
→Evidence & Trust
Source standards, provenance and transparency practices that make claims about AI systems inspectable and accountable.
→
Model intelligence dossier
Model intelligence. Measured and explained.
Structured, source-backed intelligence on model capabilities, pricing and verification status.
Browse the model index
Featured model
Claude Opus 5
- Status
- active
- Released
- Unknown — not verified
- Context window
- 1,000,000 tokens
- Max output
- 128,000 tokens
- Data cutoff
- May 2026
- Input modalities
- text and image
- Provider
- Anthropic
At a glance
Capabilities
text input, image input, text output, tool use and reasoning mode.
Pricing
$5 input · $25 output per 1,000,000 tokens.
Evidence
Official documentation, verified .
General purpose
GPT-5.6 Sol
OpenAIactive
- Context window
- 1,050,000 tokens
- Modalities
- text and image
- Pricing
- $4 input · $20 output / 1,000,000 tokens
Multimodal
Gemini 2.5 Pro
Googleactive
- Context window
- 1,048,576 tokens
- Modalities
- text, image, audio and video
- Pricing
- See full record
Reasoning model
Grok 4.6
xAIactive
- Context window
- 500,000 tokens
- Modalities
- text
- Pricing
- $2 input · $6 output / 1,000,000 tokens
Our evidence standard
Known
The value is published by a source we cite on the page.
Estimated
Derived from available evidence, shown with the method that produced it.
Not disclosed by the provider
The provider does not publish this value, so neither do we.
Unknown — not verified
We have not verified the value. We say so rather than assert one.
Not applicable
The field does not apply to this entity.
Glossary spotlight
Retrieval-Augmented Generation (RAG)
Retrieval-augmented generation (RAG) is an architecture in which a generative language model is combined with a retrieval component, so that text is generated conditioned on documents fetched from an external corpus at query time rather than on the model's parameters alone.
Explore the definition