Unverified50% confidenceBenchmarkExact time
After switching to the new architecture, Legion's Prompt dropped from 8.01M to 5.51M (-31%), Reasoning from 76.1K to 58.2K (-24%), and Completion from 62.2K to 35.8K (-42%).
1
Sources
50%
Confidence
Medium-term (~90 days)
Relevance
10/12/2026
First Seen
Valid until: 1/10/2027
Sources
Related Entities
Related Claims
UnverifiedThe reference model Supra-50M has 51.8M parameters, is 103.6 MB in bf16, 56.2 MB after 8-bit quantization, perplexity degrades from 165 to 193 when reduced to int4, and collapses entirely when compressed to ternary.65% similarUnverifiedDynacat 3.0.0's network I/O is 389 kB / 97.1 kB, a significant decrease compared to version 2.4.0's 5.03 MB / 5.95 MB.64% similarUnverified13B–14B class models (4-bit quantization) see speeds drop to 2–5 tokens per second under this configuration63% similarUnverifiedAfter vector truncation, the length is no longer 1 (the length of the first 256 dimensions is approximately 0.709), so renormalization is required.62% similarUnverifiedWhen the MTP draft depth is raised from the default 4 to 6 or 8, Qwen3.8-27B's output speed drops to 53.9 and 52.4 tokens/second respectively, below the default 55.1 tokens/second.62% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1011779API
curl https://kongchang.com/api/v1/knowledge/claims/1011779MCP
get_claim(id=1011779)