Vendor Assessment Status

Current independent assessment status of major AI model providers

Factual record of which providers have a current independent assessment. Updated with each edition of the Hastings Report.

Provider assessment status
Provider Models Status Edition
Anthropic Claude Haiku 4.5, Claude Sonnet 4 Assessed Edition 1 · April 2026
OpenAI GPT-4o Scheduled Edition 2 · May 2026
Google Gemini Pro Scheduled Edition 2 · May 2026
Meta Llama 3 70B Scheduled Edition 2 · May 2026
Mistral Mistral Large Scheduled Edition 2 · May 2026
Cohere Command R+ Scheduled Edition 2 · May 2026

GPT-2 (124M) and distilgpt2 (82M) were assessed in Edition 1 as baseline reference models. They are not listed here as they are not current production models.

What this page is and is not

This page records which providers have a current independent assessment in the Hastings Report. It is a factual status record, not a judgement. "Assessed" means the provider's model has been tested under the current methodology. "Scheduled" means assessment is planned for the next edition. No status listed means no assessment has been conducted or scheduled.

This page does not rank providers, does not imply that unassessed providers are unsafe, and does not substitute for the provider's own documentation or governance practices. The assessment tests implementation outputs under controlled conditions only.

Providers who wish to submit their models for assessment or request a vendor license should contact lab@modulithlab.com.