TRUST REPORTS

How trustworthy is the model you are about to deploy?

Eleven frontier models, each measured by Vijil Diamond on its own and again behind Vijil Dome. Every report carries the Trust Score across reliability, security and safety, the findings behind it, where the model stands among its peers, and what to fix before you deploy it. Measured September 2026; every report names the evaluation id that reproduces it.

Cover of the Vijil Trust Report for Muse Spark 1.3Muse Spark 1.3MetaREAD THE REPORT ↗Cover of the Vijil Trust Report for Claude Fable 5.1Claude Fable 5.1AnthropicREAD THE REPORT ↗Cover of the Vijil Trust Report for Qwen 3.8 MaxQwen 3.8 MaxAlibabaREAD THE REPORT ↗Cover of the Vijil Trust Report for DeepSeek v4 ProDeepSeek v4 ProDeepSeekREAD THE REPORT ↗Cover of the Vijil Trust Report for Grok 4.6Grok 4.6xAIREAD THE REPORT ↗Cover of the Vijil Trust Report for Kimi K3Kimi K3Moonshot AIREAD THE REPORT ↗Cover of the Vijil Trust Report for GLM 5.3GLM 5.3Z.aiREAD THE REPORT ↗Cover of the Vijil Trust Report for GPT 5.6 SolGPT 5.6 SolOpenAIREAD THE REPORT ↗Cover of the Vijil Trust Report for Gemini 3.8 FlashGemini 3.8 FlashGoogleREAD THE REPORT ↗Cover of the Vijil Trust Report for Mistral Medium 3.5Mistral Medium 3.5Mistral AIREAD THE REPORT ↗Cover of the Vijil Trust Report for Mistral Large 3Mistral Large 3Mistral AIREAD THE REPORT ↗
Get the reports
All eleven, free. Tell us who you are and they open straight away.
Thanks — every report is open now. Pick any one.
By submitting, you agree to Vijil's privacy policy. We’ll only use your details to send you these reports.