GDM-MRCR v2
7 models tested · Updated 2026-07-21 · Verified sources only
Gemini 3.6 Flash leads at 91.8%
1
Google DeepMind · Blog/Google DeepMind · 2026-07-21
128K 8-needle MRCR. Beats Gemini 3.1 Pro (84.9) and GPT-5.6 (74.8).
91.8%
2
Google DeepMind · Blog/Google DeepMind (cross-ref) · 2026-07-21
Google official 128K 8-needle MRCR.
84.9%
3
Google's measurement of Grok 4.5 on 128K MRCR.
81.4%
4
Google · Blog/Google DeepMind · 2026-07-21
128k average. Normalized from MRCR v2 to GDM-MRCR v2 (same benchmark, Google DeepMind naming).
77.3%
5
OpenAI · Blog/Google DeepMind (cross-ref) · 2026-07-21
Google's measurement of GPT-5.6 Luna 128K 8-needle MRCR.
74.8%
6
Google DeepMind · Blog/Google DeepMind · 2026-07-21
Beats 3.1 Flash-Lite (60.1%) on 8-needle MRCR.
72.2%
7
Anthropic · Blog/Google DeepMind (cross-ref) · 2026-07-21
Google's measurement of Sonnet 5 128K MRCR.
71.6%