Yesterday
-
Revert card to 11 Nov text (undo 88a5e11d, 2875facc)
-
Card: keep only our own Artificial Analysis number
-
Card: date the Artificial Analysis result, drop the stale ranking table
7 days ago
-
Card: state the 4K option gap in percentage points
-
Card: Kapitsa and MLX speed at 1K, 4K and 16K
9 days ago
-
Add paper link (#1)

ozyman +1: nielsr - Add paper link (6f337cbaa065271ba8cf1be2c5425787e4308530)
Co-authored-by: Niels Rogge <nielsr@users.noreply.huggingface.co>
19 days ago
-
Cite the results datasets on their resolving pages
-
Name the Hugging Face CLI before hf download
-
Card: affine Q4 runs on 24 GB Macs
-
Card: affine Q4 Kapitsa command
-
Card: Kapitsa serves the affine Q4 folder
-
Card: Kapitsa imports the Q8_0 GGUF
-
Card: add Kapitsa and MLX setup
-
Card: link the affine Q4 checkpoint
-
Card: drop base_model metadata
-
Model card: Artificial Analysis Decision Index results; describe backbone as a 27B model
21 days ago
-
Label the public index as Decision Index 0.4. Scores unchanged.
-
Put JevBench 92.2% (213/231) and Decision Bench v4 94.30% (1010/1071) back on the card.
-
Correct the card to Decision Index 0.3 results (67.52 / vision 73.92). Remove benchmarks.png.
30 days ago
-
Add benchmark table image
-
Model card: Decision Index 0.3.2
-
Model card: refer to the model as Drex Max
-
Model card: llama.cpp and Ollama sections, Q8_0 GGUF link
-
Release card: canonical license, 16,384 default context, links to nace-ai/drex-decision-models; public flag removed from card
-
Duplicate from NischayDnk/Drex-Max