Introduction

The Gemini 3 Pro model of Google has debuted recently and has achieved a record score of 69% trust in the blinded test by Prolific, surpassing its predecessor Gemini 2.5 with a 53% increase.

Technical details

The model was evaluated in a blind test of 26,000 users, which assessed the performance of the model in different scenarios.

The test measured user trust, adaptability and communication style.

Practical implications

The result of the test highlights the importance of using neutral and objective evaluation methods to determine the performance of natural language models.

Prolific created a benchmark called HUMAINE that applies this approach.

Conclusion

The result shows that Gemini 3 Pro is the most reliable and secure model for use in various situations.

Prolific hopes to continue improving its evaluation methods to ensure the quality of natural language models.

\n