KronLSTM Study - Language - ARC-Challenge: leaderboard
KronLSTM (2026): matched research-trained 1.5B models on 100B pretraining tokens, linear and hybrid architectures kept distinct; zero/few-shot downstream task protocol from the paper. Architecture names do not identify unrelated public Mamba, mLSTM or GDN checkpoints. Accuracy (%); higher is better.
Metric: Accuracy (%). Source: arxiv.org. Saturation forecast: Rough model projection: around 2026. 12 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | KronLSTM Study KronLSTM 1.5B Hybrid (100B tokens) | 44.11 | |
| 2 | KronLSTM Study KronLSTM 1.5B Linear (100B tokens) | 43.6 | |
| 3 | KronLSTM Study Mamba-3 1.5B Hybrid (100B tokens) | 43.09 | |
| 4 | KronLSTM Study GDN-2 1.5B Hybrid (100B tokens) | 42.75 | |
| 5 | KronLSTM Study GDN 1.5B Hybrid (100B tokens) | 42.58 | |
| 6 | KronLSTM Study mLSTM 1.5B Hybrid (100B tokens) | 42.58 | |
| 7 | KronLSTM Study GDN-2 1.5B Linear (100B tokens) | 41.89 | |
| 8 | KronLSTM Study KDA 1.5B Linear (100B tokens) | 41.55 | |
| 9 | KronLSTM Study mLSTM 1.5B Linear (100B tokens) | 41.3 | |
| 10 | KronLSTM Study KDA 1.5B Hybrid (100B tokens) | 41.21 |
Interactive version: theaggregate.ai/benchmark?slug=kronlstm-study-language-arc-challenge · How It Works · Data refreshed daily, snapshot 2026-10-21.