| Dataset Name | SOTA Method | Metric | Trend | ||
|---|---|---|---|---|---|
| LongBench v2 | Qwen-3.5 | Average Score68.9 | 113 | 22d ago | |
| LongBench | GHG-TDA | Score73.8 | 107 | 24d ago | |
| LongBench | CAP-CoT | Accuracy (LongBench)70.4 | 101 | 2mo ago | |
| LoCoMo | FluxMem | F1 (Multi Hop)93.26 | 78 | 1mo ago | |
| BABILong 16k | Accuracy29.8 | 72 | 3mo ago | ||
| BABILong 8k | Accuracy34.7 | 65 | 3mo ago | ||
| BABILong 4k | Accuracy (BABILong 4k)38.5 | 51 | 3mo ago | ||
| InfiniteBench | Overall Score36.51 | 45 | 24d ago | ||
| OOLONG | λ-RLM | Accuracy68.4 | 37 | 3mo ago | |
| LongReason Average across 16k-64k | Accuracy54.2 | 36 | 29d ago | ||
| LongReason 64k prefill length | Accuracy53.02 | 36 | 29d ago | ||
| LongReason 32k prefill length | Accuracy53.9 | 36 | 29d ago | ||
| LongReason 16k prefill length 1.0 (test) | Accuracy55.67 | 36 | 29d ago | ||
| Long-context Benchmarks 100K context LB-V2 DocMath Frames LB-MQA (test) | Qwen3-30B-A3B-Thinking + SPELL | DocMath Score66.7 | 36 | 4mo ago | |
| Long-context Benchmarks 16K context DocMath Frames LB-MQA V2 (test) | Qwen3-30B-A3B-Thinking + SPELL | DocMath64.1 | 36 | 4mo ago | |
| RULER | HyLo-Llama-14MLA14M2 | RULER Performance (8K Context)75.3 | 35 | 2mo ago | |
| AA-LCR | Kimi-K2.6 | Score70.2 | 35 | 1mo ago | |
| LongReason 64K-input 70K context | KVZip | Accuracy71.25 | 34 | 1mo ago | |
| ∞ Bench | MiA (Emb-Only) | Accuracy90.39 | 32 | 4mo ago | |
| OOLONG trec_coarse | Kimi K2 | Score86.6 | 28 | 4mo ago | |
| FRAMES | Score84.7 | 27 | 23d ago | ||
| OOL-Pairs | Latency (s)5.1 | 27 | 4mo ago | ||
| OOLONG | Latency (s)7.1 | 27 | 4mo ago | ||
| LongGenBench 8K | GSM8K Score44.51 | 22 | 1mo ago | ||
| LongGenBench 4K | GSM8K Score53.18 | 22 | 1mo ago |
