Skip to content

MRCRv2

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

12 models

Higher is better.

  1. 1Qwen3.8 MaxAlibaba92.9
  2. 2Qwen3.7 PlusAlibaba91.7
  3. 3Qwen3.7 MaxAlibaba90.4
  4. 4Gemini 3.5 FlashGoogle77.3
  5. 5Mercury 3Inception75.5
  6. 6Gemini 3.5 Flash-LiteGoogle72.2
  7. 7Gemma 4 31BGoogle66.4
  8. 8Pokee-Isaac 28BPokee AI60.7
  9. 9Gemma 4 26B A4BGoogle44.1
  10. 10Gemma 4 12BGoogle43.4
  11. 11Gemma 4 E4BGoogle25.4
  12. 12Gemma 4 E2BGoogle19.1
Higher is better. Bar colour: blue proprietary, orange open weights, green other.

About this benchmark

Metric
Multi-round long-context evaluation
Tasks
Long-context retrieval
Authors
OpenAI (2025)

Scores are as reported by model developers or evaluators and may use different settings (for example reasoning effort or tools). Compare with care.