# MRCR 1M leaderboard

Reasoning benchmark

> Every figure below is reproduced as its upstream source published it: nothing is modelled, estimated, interpolated or converted. `Unavailable` means no source published the value — it is never a zero. Each value carries its evidence state and the date it was observed.

Page: https://modelchoice.dev/leaderboards/benchmarks/reasoning-mrcr1m

Every model the catalog carries a published MRCR 1M value for, ranked by that value.

## Benchmark record

| Field | Value |
| --- | --- |
| Category | Reasoning |
| Measure | Long-context retrieval MMR |
| Tasks | Million-token retrieval |
| Difficulty | Million-token long context |
| Published by | [DeepSeek-V4 Technical Report](https://huggingface.co/deepseek-ai/DeepSeek-V4-Pro/blob/main/DeepSeek_V4.pdf) |

A million-token MRCR long-context retrieval benchmark reported in DeepSeek-V4 model evaluations.

## MRCR 1M ranking

5 models with a published MRCR 1M value, ordered by that value, highest first. Models the source has not scored on this benchmark are not listed — they are unmeasured, not last.

| Rank | Model | Provider | Long-context retrieval MMR | Evidence |
| --- | --- | --- | --- | --- |
| 1 | [DeepSeek V4 Pro](https://modelchoice.dev/models/deepseek-v4-pro) | DeepSeek | 83.5 | Observed 2026-11-21 · source benchlm:models |
| 1 | [DeepSeek V4 Pro 0813](https://modelchoice.dev/models/deepseek-v4-pro-max) | DeepSeek | 83.5 | Last good 2026-11-18 · source benchlm:models |
| 3 | [DeepSeek V4 Flash 0731](https://modelchoice.dev/models/deepseek-v4-flash-max) | DeepSeek | 78.7 | Observed 2026-11-21 · source benchlm:models |
| 4 | [Muse Spark 1.1](https://modelchoice.dev/models/muse-spark-1-1) | Meta | 54.1 | Observed 2026-11-21 · source benchlm:models |
| 5 | [Gemini 3.5 Flash](https://modelchoice.dev/models/gemini-3-5-flash) | Google | 26.6 | Observed 2026-11-21 · source benchlm:models |

Rows are ordered by the value the source published, highest first. The source does not state whether a higher value is the better result, so this document does not either: for a benchmark that measures a rate of failure, read the table from the bottom.
