MRCRBrowse 296

MRCR

Can the model retain and reason over evidence spread across a very long input?

Version not specifiedExact reported variant
Long Context15 ranked models21 reported values4 reportsHigher is better

Top rankings

One row per model, using its best reported score across effort settings.

15
RankModelBest scoreBest reported setting
1Gemini 3.7 FlashGoogle97.0%Reported configurationGoogle report1 value · 1 reportAug 13, 2026 · source
2GPT-5.5OpenAI94.8%Reported configurationGoogle report1 value · 1 reportMay 19, 2026 · source
3GPT-5.6 TerraOpenAI93.5%Reported configurationGoogle report1 value · 1 reportAug 13, 2026 · source
4Gemini 3.6 FlashGoogle91.8%Reported configurationGoogle report2 values · 2 reportsAug 13, 2026 · source
5Gemini 3.1 ProGoogle84.9%Reported configurationGoogle report3 values · 3 reportsJul 21, 2026 · source
6Claude Sonnet 4.6Anthropic84.9%Reported configurationGoogle report2 values · 2 reportsMay 19, 2026 · source
7Claude Opus 4.6Anthropic84.0%max effortGoogle report1 value · 1 reportFeb 19, 2026 · source
8GPT-5.2OpenAI83.8%xhigh effortGoogle report1 value · 1 reportFeb 19, 2026 · source
9Claude Sonnet 5Anthropic81.5%Reported configurationGoogle report2 values · 2 reportsAug 13, 2026 · source
10Grok 4.5SpaceXAI81.4%Reported configurationGoogle report1 value · 1 reportJul 21, 2026 · source
11Gemini 3.5 FlashGoogle77.3%Reported configurationGoogle report2 values · 2 reportsJul 21, 2026 · source
12Gemini 3 ProGoogle77.0%high effortGoogle report1 value · 1 reportFeb 19, 2026 · source

Effort curve

Every sourced cost-linked effort value for this exact version. Lines connect complete sweeps only.

0
No cost-linked effort sweep for this version.
Definition and comparison boundarypublic methodology

Can the model retain and reason over evidence spread across a very long input?

Long-context retrieval and reasoning, not context-window size alone. Higher is better. The value is the percentage reported in this lab's table.

Harness, tools, prompts, attempt count, and benchmark version can materially change the result. Compare within one reporting context.

Google-computed cumulative 128K score.

Method / source