Today
-
card: measured Thor numbers (2026-10-30: 25.0 tok/s decode, 739 tok/s prefill), CUDA server repo + start.sh, NVFP4 overlay
-
add the NVFP4 dense overlay (separate copy; set DSV4_NVFP4_OVERLAY to serve it)
-
add tokenizer_config.json (from the base model) so v41_server needs no source checkpoint
-
add tokenizer.json (from the base model) so v41_server needs no source checkpoint
4 days ago
-
docs(card): add inline provenance citations, verification ledger, and sharded GGUF/128GB memory breakdown
-
docs(card): upgrade model card with SOTA HF metadata, summary tables, and sibling links
-
s12: model card generated from s10 results
-
s12: reap_keepset.json
-
s12: thor.json
-
s12: vision.safetensors
-
s12: mtp-2.safetensors
-
s12: mtp-1.safetensors
-
s12: mtp-0.safetensors
-
s12: layer-39.safetensors
-
s12: layer-38.safetensors
-
s12: layer-37.safetensors
-
s12: layer-36.safetensors
-
s12: layer-35.safetensors
-
s12: layer-34.safetensors
-
s12: layer-33.safetensors
-
s12: layer-32.safetensors
-
s12: layer-31.safetensors
-
s12: layer-30.safetensors
-
s12: layer-29.safetensors
-
s12: layer-28.safetensors
-
s12: layer-27.safetensors
-
s12: layer-26.safetensors
-
s12: layer-25.safetensors
-
s12: layer-24.safetensors
-
s12: layer-23.safetensors
-
s12: layer-22.safetensors
-
s12: layer-21.safetensors
-
s12: layer-20.safetensors
-
s12: layer-19.safetensors
-
s12: layer-18.safetensors
-
s12: layer-17.safetensors
-
s12: layer-16.safetensors
-
s12: layer-15.safetensors
-
s12: layer-14.safetensors
-
s12: layer-13.safetensors
-
s12: layer-12.safetensors
-
s12: layer-11.safetensors
-
s12: layer-10.safetensors
-
s12: layer-09.safetensors
-
s12: layer-08.safetensors
-
s12: layer-07.safetensors
-
s12: layer-06.safetensors
-
s12: layer-05.safetensors
-
s12: layer-04.safetensors
-
s12: layer-03.safetensors
