Filters
Search filters
Search the tags or the topic text.
Present
Future
Both
Filter tags
Models
Tokenizers
Research
Current Runs
Bugs
New Ideas
Topics
Pick one topic to open the detail page.
StenToken Progress report: `english_code` is at `98 GB/150.0 GB`, and the main phase is nearing its finish line. Last updated 2026-06-30 Tokenizers · Current Runs · Research · New Ideas
StenToken-FineWeb Planned StenToken variant trained on 100% FineWeb for dataset-specific compression. Last updated 2026-07-01 Tokenizers · Research · New Ideas
StenToken-FineWeb-Edu Planned StenToken variant trained on 100% FineWeb-Edu for dataset-specific compression. Last updated 2026-07-01 Tokenizers · Research · New Ideas
Stentor4-80M Progress report: the current Stentor4 build drops z-loss, removes embedding weight decay, and turns off QK-normalization. Last updated 2026-06-30 Models · Current Runs · Bugs · Research · New Ideas
Stentor4-Preview Preview track for Stentor4 using the StenToken schedule so followers can see the model sooner instead of waiting for a separate 32K-tokenizer path. Last updated 2026-06-30 Models · Current Runs · Research · New Ideas
SLM ARENA COMMING SOON!!! The arena code is finished and is now just being debugged. Last updated 2026-07-01 Models · Research · New Ideas
Archived