4 days ago
-
Benchmark page: Qwen3.6-35B-A3B, Qwen 3.8 DWQ, gpt-oss-20b, Qwen3 Reranker, Arete 27B, Axion 1.0 35B
9 days ago
-
Benchmark page: Qwen 3.8 27B draft-ahead decoding (test setup; off in the app for now)
10 days ago
-
Benchmark page: Smarty draft-ahead decoding on 500 GSM8K problems
-
Benchmark page: Smarty DWQ 3-bit/4-bit vs served build on 500 GSM8K problems
11 days ago
-
Benchmark page: base model then improved versions; 30-problem weight-edit reproduction (MIT CSAIL p50c)
