Rumors: New Gemini 3.5 Checkpoint Impresses and Beats Opus 5 Max

Recently, tests of Google's next-gen Gemini 3.5 (and potentially 3.6) models have sparked heated discussions on Reddit. Multiple users reported that the new checkpoint being tested on the Arena platform is incredibly impressive, with rumors even suggesting it has defeated the Claude Opus 5 Max Thinking model.

Confirmed

Current information primarily stems from early hands-on feedback by community users:

1. **Stunning Performance**: User @Last_Conclusion_8984 noted that the new checkpoint of Gemini 3.5 Pro is "ridiculously strong" and provided a demo video.

2. **Improved "Laziness"**: User @Able-Line2683 mentioned that early tests show this version no longer exhibits the "lazy" tendencies during tasks that were present in Gemini 3.1 Pro.

Unconfirmed

1. **Defeating Competitors**: Reddit rumors relayed by @Last_Conclusion_8984 claim this checkpoint beat Claude Opus 5 Max Thinking in tests. However, the UI in the provided screenshot displayed the model name as "gemini-3.6-flash", meaning the exact version number and actual capabilities still need verification.

2. **Lack of Rigorous Data**: All current evaluations are based on subjective experiences and demo videos. There are no authoritative Benchmark results yet to confirm these astonishing performance leaps.

Why It Matters

Although these are currently just community rumors, the dense positive feedback sends a strong signal. If the reports regarding performance improvements and the resolution of the "laziness" issue are true, it means Google's next-generation LLM will be highly competitive. It remains worth closely monitoring for subsequent official releases and benchmark evaluations.

2026-07-26 ~ 2026-07-27 · 5 related posts

Primary sources