Skip to content

OpenAI MRCR v2 8-needle 512K-1M

MRCR v2 slice focused on retrieval across 512K-1M-token contexts.

2 models

Higher is better.

  1. 1Muse Spark 1.3Meta98.1
  2. 2GPT-6 AstraOpenAI96.3
Higher is better. Bar colour: blue proprietary, orange open weights, green other.

About this benchmark

Metric
Mean sequence-matcher ratio
Tasks
100 eight-needle retrieval examples
Authors
Meta Superintelligence Labs (2026)

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