OpenAI-MRCR: 2 needle 128k
Multi-round Co-reference Resolution (MRCR) benchmark for evaluating an LLM's ability to distinguish between multiple needles hidden in long context. Models are given a long, multi-turn synthetic conversation and must retrieve a specific instance of a repeated request, requiring reasoning and disambiguation skills beyond simple retrieval.
9rows
scoreprimary metric
2026-05-06sampled
Metadata
Metrics
Score, Normalized Score
| Rank | Subject | Score | Model Match | Provenance | Sampled |
|---|---|---|---|---|---|
| 1 | GPT-5 | 0.95 |  GPT-5 openai-gpt-5 | Self-reported | 2026-05-06 |
| 2 | MiniMax M1 40K | 0.76 | — | Self-reported | 2026-05-06 |
| 3 | MiniMax M1 80K | 0.73 | — | Self-reported | 2026-05-06 |
| 4 | GPT-4.1 | 0.57 | GPT-4.1 openai-gpt-4.1 | Self-reported | 2026-05-06 |
| 5 | GPT-4.1 mini | 0.47 | GPT-4.1 Mini openai-gpt-4.1-mini | Self-reported | 2026-05-06 |
| 6 | GPT-4.5 | 0.39 | GPT-4.5 openai-gpt-4.5-preview | Self-reported | 2026-05-06 |
| 7 | GPT-4.1 nano | 0.37 | GPT-4.1 Nano openai-gpt-4.1-nano | Self-reported | 2026-05-06 |
| 8 | GPT-4o | 0.32 | GPT-4o (2024-08-06) openai-gpt-4o-2024-08-06 | Self-reported | 2026-05-06 |
| 9 | o3-mini | 0.19 | o3-mini openai-o3-mini | Self-reported | 2026-05-06 |
No matching rows.