Unverified50% confidenceFactExact time
Betelgeuse strips the expert payload from Shard 1, making the main files only 3.2GB (Dense.gguf) + 35.3GB (experts.bin, 4KB-aligned single read blocks) + 26.8GB (PLE.gguf n-gram shards), avoiding keeping 35GB of dead expert tensors in each of the Shard 1 .gguf and experts.bin.
1
Sources
50%
Confidence
Medium-term (~90 days)
Relevance
10/31/2026
First Seen
Valid until: 1/29/2027
Sources
Running a 68GB Large Model on a 12GB GPU: Betelgeuse Engine Benchmarked at 22 tok/s
redditr/LocalLLaMA10/31/2026
Related Entities
Related Claims
UnverifiedFor ModelFit on Kimi K3 (unsloth repository), the UD-Q1_0 quantization has a measured file size of 466.4 GB and requires 561.1 GB of VRAM/RAM at an 8K context; UD-Q4_K_XL has a measured file size of 1.51 TB and requires about 1.6 TB; F16 is about 3 TB.70% similarUnverifiedThe HGN format's PLE module uses FP8 quantization at approximately 47.7 GB, while the GGUF format quantizes PLE to Q4 via unsloth at only 26.8 GB, but the overall file is larger and occupies approximately 10 GB of additional VRAM68% similarUnverifiedGoRoute uses less than 6MB of memory when idle, and the compiled single binary is 16MB in size.67% similarUnverifiedIn Unsloth v0.1.913-beta, peak memory when previewing a 1 GB local Parquet file fell from 1.5 GB to about 300 MB.66% similarUnverifiedThe model has 1.91M parameters, and after int8 quantization the weights occupy 1.97 MB, which is 49.8% of the Ryzen 5 3400G's 4 MB L3 cache.66% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1120769API
curl https://kongchang.com/api/v1/knowledge/claims/1120769MCP
get_claim(id=1120769)