Testing AI on Real-World Tasks

GPT-5.6 Sol evaluated across the Pramaana benchmark suite

The model places second on the Pramaana Index and leads on agentic coding, cyber security, and legal research evaluations.

• All runs use immutable model and benchmark configurations.
• Published scores include uncertainty, cost, latency, and source provenance.
• Results are independently reviewed before they enter public snapshots.