AI benchmarks and leaderboards are different things
Learn the difference between a benchmark, an evaluation run and a leaderboard, and use each to answer the right question.
Model reports
MRCR
Needle position and generation seed change the result.
Evaluation context and evidence
- Source
- Model reports
- Publisher
- Not recorded
- Benchmark release
- 2025-04-14
- Collected
- Not recorded
- Modality
- Not recorded
These are source-reported observations, not independent reruns by this site. Check the task version, evaluation protocol, unit and reporting date before comparing models. Results from different setups are not a universal ranking.
Original sources
Reported results 7
Shown in source observation order, not ranked. Dates retain the precision provided by the source.
| Model | Value | Unit | Protocol | Reported date and basis | Source |
|---|---|---|---|---|---|
| DeepSeek-V4-Pro | 44.7 | percent | non-think, MMR | 2026-04-22 · document_publication | DeepSeek-V4-Pro |
| DeepSeek-V4-Pro | 83.3 | percent | think high, MMR | 2026-04-22 · document_publication | DeepSeek-V4-Pro |
| DeepSeek-V4-Pro | 83.5 | percent | think max, MMR | 2026-04-22 · document_publication | DeepSeek-V4-Pro |
| DeepSeek-V4-Flash | 78.7 | percent | think max, MMR | 2026-06-25 · document_publication | DeepSeek-V4 |
| Gemini 3.1 Pro | 84.9 | percent | Thinking (High), 128k (average) | 2026-02-19 · document_publication | Gemini 3.1 Pro |
| Gemini 3.1 Pro | 26.3 | percent | Thinking (High), 1M (pointwise) | 2026-02-19 · document_publication | Gemini 3.1 Pro |
| Gemma 4 (31B) | 66.4 | percent | 8-needle, 128k (average) | 2026-03-11 · document_publication | Gemma 4 (31B) |