# OpenAI MRCR v2 8-needle 512K-1M (MRCR v2 512K-1M)

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

Canonical page: https://benchlm.ai/benchmarks/mrcr-v2-512k-1m

- Category: [Reasoning](/reasoning)
- Last updated: October 17, 2026

## About MRCR v2 512K-1M

- Year: 2026
- Tasks: 100 eight-needle retrieval examples
- Format: Mean sequence-matcher ratio
- Difficulty: Million-token retrieval
- Paper: [Muse Spark 1.3 Evaluation Methodology](https://research.meta.ai/static/muse-spark-1-3-multimodal-evaluation-methodology)

Meta re-bins OpenAI's MRCR v2 data by o200k_base token count and evaluates 100 examples with eight needles. The mean sequence-matcher ratio stays display-only because this slice is not part of the weighted reasoning schema.

MRCR v2 512K-1M is currently displayed on BenchLM for reference, but it is excluded from the weighted scoring formula.

## Leaderboard (2 models)

| Rank | Model | Creator | Score |
|------|-------|---------|-------|
| 1 | [Muse Spark 1.3](/models/muse-spark-1-3) | Meta | 98.1% |
| 2 | [GPT-6 Astra](/models/gpt-6-astra) | OpenAI | 96.3% |

## FAQ

### What does MRCR v2 512K-1M measure?

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

### Which model scores highest on MRCR v2 512K-1M?

Muse Spark 1.3 by Meta currently leads with a score of 98.1% on MRCR v2 512K-1M.

### How many models are evaluated on MRCR v2 512K-1M?

2 AI models have published results on MRCR v2 512K-1M in the BenchLM catalog.

### Does MRCR v2 512K-1M affect BenchLM's overall score?

Not directly. MRCR v2 512K-1M is still displayed on BenchLM for reference, but it is excluded from the weighted scoring formula.

## Compare Top Models on MRCR v2 512K-1M

- [Muse Spark 1.3 vs GPT-6 Astra](/compare/gpt-6-astra-vs-muse-spark-1-3)