OpenAI MRCR
K-EXAONE 2.0 94.4% leads since Aug 2026 Dropped Models read a long synthetic chat that hides 2, 4 or 8 identical writing requests and must reproduce the response to the i-th one. Scored by mean string match ratio, in context bins from 4K to 1M tokens.
Inspired by Google's MRCR but built from new GPT-4o text, so it is a different dataset from GDM MRCR v2. Labs report it by needle count and context bin.
- Level
- Everyday Most adults, without training
- Scored
- Checked automatically A program checks the result against an answer key, unit tests or the expected final state.
- Answers
- Short answer A number, word or short string matched to a key.
K-EXAONE 2.0 leads with 94.4% among models released by Jul 2026
Best score per model as labs report them in their own cards. Labs pick their own settings, so these compare less cleanly than independent runs.
The newest model with a score here came out in Jul 2026, and nobody has reported a newer one. Among the 20 highest-ranked models on the capability index, these have no score yet:
- GPT-6 Astra
- Claude Opus 5.5
- GPT-6.1 Terra
- GPT-6.1 Sol
- and 16 more
- 1 K-EXAONE 2.0 94.4%
- 2 GPT-5 mini 84.3%
- 3
MiniMax-M1-80k 73.4% - 4 GPT-4.1 57.2%
- 5 Kimi K2 Jul 2025 55%
- 6 K-EXAONE 52.3%
- 7 GPT-4.1 mini 47.2%
- 8 GPT-5 nano 43.2%
- 9 GPT-4.1 nano 36.6%
- 10
Step 3.5 Flash 28.8%
| # | Model | Score | Setting | Date | Source |
|---|---|---|---|---|---|
| 1 | K-EXAONE 2.0 | 94.4% | Aug 2026 | Model card | |
| 2 | GPT-5 mini | 84.3% | Aug 2025 | Model card | |
| 3 | MiniMax-M1-80k | 73.4% | Jun 2025 | Model card | |
| 4 | GPT-4.1 | 57.2% | Apr 2025 | Model card | |
| 5 | Kimi K2 Jul 2025 | 55% | Jul 2025 | Model card | |
| 6 | K-EXAONE | 52.3% | Jan 2026 | Model card | |
| 7 | GPT-4.1 mini | 47.2% | Apr 2025 | Model card | |
| 8 | GPT-5 nano | 43.2% | Aug 2025 | Model card | |
| 9 | GPT-4.1 nano | 36.6% | Apr 2025 | Model card | |
| 10 | Step 3.5 Flash | 28.8% | Feb 2026 | Model card |
4 models have held number one since its release in Apr 2025
Lab claims count from the day the card came out, independent runs from the model’s release.
The best score passed 85% in Aug 2026
All the scores we have, one dot each, so a model tested at several settings appears more than once.
Labs moved on to OpenAI MRCR v2
Launch posts from every lab that has listed it, one tick per post. Tall colored ticks include it, and short gray ones leave it out.
| Date | Launch post | Score | As the post writes it |
|---|---|---|---|
| 7 Aug 2025 | GPT-5 nano | 43.2 | OpenAI-MRCR: 2 needle 128k |
| 7 Aug 2025 | GPT-5 mini | 84.3 | OpenAI-MRCR: 2 needle 128k |
| 16 Jun 2025 | MiniMax-M1-80k | 73.4 | Best of OpenAI-MRCR (128k) / OpenAI-MRCR (1M) |
| 14 Apr 2025 | GPT-4.1 nano | 36.6 | OpenAI-MRCR: 2 needle 128k 36.6 (1M: 12.0) |
| 14 Apr 2025 | GPT-4.1 mini | 47.2 | OpenAI-MRCR: 2 needle 128k 47.2 (1M: 33.3) |
| 14 Apr 2025 | GPT-4.1 | 57.2 | Best of OpenAI-MRCR: 2 needle 128k / OpenAI-MRCR: 2 needle 1M |
7 system cards report it, from 5 labs
Besides launch posts, we read the system cards, model cards and technical reports labs publish with a release. They report far more, including safety and frontier-risk tests.
| Date | Document | Score | As the document writes it | Section |
|---|---|---|---|---|
| 5 Aug 2026 | K-EXAONE 2.0 technical report | 94.4 | OpenAI-MRCR (Table 6, Reasoning mode) | Capabilities |
| 11 Feb 2026 | Step 3.5 Flash technical report | 28.8 | MRCR-8needle, AUC@128k (Table 5 and E.2.5). PaCoRe 26.3 | Capabilities |
| 5 Jan 2026 | K-EXAONE technical report | 52.3 | OpenAI-MRCR, 2, 4 and 8 needle bins up to 128K (Table 3, Reasoning mode and C.8). Non-reasoning 60.9 in Table 4, the text says 65.9 | Capabilities |
| 2 Dec 2025 |  Nova 2 Lite technical report | OpenAI-MRCR (8-needle), 8K to 1M line chart without printed values (Figure 4) | Capabilities | |
| 2 Dec 2025 | Nova 2.0 Pro Preview technical report | OpenAI-MRCR (8-needle), 8K to 1M line chart without printed values (Figure 4) | Capabilities | |
| 21 Jul 2025 | Kimi K2 technical report | 55 | MRCR (Acc.), OpenAI MRCR dataset, 128K window (Table 3) | Capabilities |
| 16 Jun 2025 | MiniMax-M1-80k technical report | 73.4 | OpenAI-MRCR (128k), shown as MRCR (4-needle) in Figure 1 (Table 2). OpenAI-MRCR (1M) 56.2 | Capabilities |
Versions
We switch the top row to each new version when it comes out. Open it for one row per version. An old version stays in play while people still test it, so whoever first takes it past 85% gets credit for closing it out.
More in Long context
- MRCR v2 In use Multi-round coreference: retrieve the n-th of several similar requests hidden in a long synthetic conversation, up to 1M tokens.
Muse Spark 1.3 98.5% - LongBench v2 In use 503 multiple-choice questions over 8K-2M-word contexts (QA, dialogue, code repos, structured data). Accuracy. Beam 1.0 66.9%
- GraphWalks In use A graph is given as an edge list in long context. The model answers BFS or parent queries, scored by F1. Claude Fable 5 100%
- CL-bench In use Model must learn new knowledge from a supplied long context (rules, procedures, data) and solve tasks. Strict LLM-verified rubrics. GPT-5.4 27.9%
- AA-LCR 1.1 In use Models answer 100 questions that need reasoning across sets of real documents of about 100K tokens. An LLM equality checker with a fixed system prompt compares each answer with the key. Step 5 Preview 88.3%
- LOCA-Bench In use Agentic tasks whose environment state is scaled to grow context length while task semantics stay fixed. Success rate versus context length. GPT-5.2 21.3%