Inference
EN

MRCRv2

A long-context benchmark for memory, retrieval, and multi-round coherence over large contexts.

  • In index
  • Reasoning
ModelsModels with a result.14
Top resultBest result on this benchmark.93.6%Sakana Fugu-Ultra
Top-3 spreadPoints from first to third.1.9 pts
YearYear of release.—

Result and price

6 of 14 results

Results

14 results
#ModelResultThe score from the source.
01Sakana Fugu-Ultra93.6%
02Qwen3.8 Max92.9%
03Qwen3.7 Plus91.7%
04Qwen3.7 Max90.4%
05Sakana Fugu86.6%
06Gemini 3.5 Flash77.3%
07Mercury 375.5%
08Gemini 3.5 Flash-Lite72.2%
09Gemma 4 31B66.4%
10Pokee-Isaac 28B60.7%
11Gemma 4 26B A4B44.1%
12Gemma 4 12B43.4%
13Gemma 4 E4B25.4%
14Gemma 4 E2B19.1%