Kimi Linear 48B A3B Instruct: benchmark results
Moonshot's 48B MoE (3B active) with hybrid Kimi Delta linear attention and a 1M context, MIT-licensed. Provider: Moonshot. Released 2025-10-30. Access: Open.
Aggregate Rating 1436 ± 37, rank #1072 of 1656 rated models, from 31 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| BacktestBench - Indicator Retrieval | 74.37 | Accuracy (%) of the Summarizer stage in retrieving the facto | 81.8 |
| UGI - Writing | 37.74 | Writing Score | 63.6 |
| UGI - Natural Intelligence | 22.09 | NatInt Score | 50.3 |
| AA Terminal-Bench Hard | 11.36 | Accuracy (%) | 46.8 |
| AA LiveCodeBench | 37.78 | Pass@1 (%) | 44.7 |
| SpeechMap Compliance | 53.4 | % Requests Completed | 37.3 |
| AA AIME 2025 | 36.33 | Accuracy (%) | 36 |
| AA Long Context Reasoning | 28 | Accuracy (%) | 27.7 |
| Artificial Analysis Intelligence Index | 7.3 | Intelligence Index | 25.4 |
| ATLAS | 33.1 | ATLAScore (self-reported) | 22.7 |
| SvelteBench | 66.7 | Average pass@1 (%) | 22 |
| AA MMLU-Pro | 58.5 | Accuracy (%) | 21.8 |
Interactive version: theaggregate.ai/model?slug=kimi-linear-48b-a3b-instruct · How It Works · Data refreshed daily, snapshot 2026-10-20.