Unverified50% confidenceBenchmarkExact time
In code scenarios, each verification step accepts about 7.4 tokens, reaching a throughput of 154 tok/s; in prose scenarios each step accepts about 2.5 tokens, with a throughput of 35-37 tok/s; the ordinary decoding baseline throughput is about 18-19 tok/s.
1
Sources
50%
Confidence
Medium-term (~90 days)
Relevance
10/25/2026
First Seen
Valid until: 1/23/2027
Sources
Related Entities
Related Claims
UnverifiedOn the 35 prompts that the router kept local, Llama 3.2 1B's answers were judged acceptable 30 times (86%, average 4.4/5); on the 23 text prompts sent to the cloud, the same model was acceptable only 4 times (17%, average 2.0/5).73% similarUnverifiedCode-type tasks can achieve a throughput improvement of about 4x or more (154 vs 18-19 tok/s), while prose-type tasks achieve only about 2x (35-37 vs 18-19 tok/s).72% similarUnverifiedAfter adopting the causal encoder-decoder, the activated parameters per token during the prefill phase were reduced from 16B to 8B.72% similarUnverifiedBased on measurements from more than 7,500 real API counts, the median call for claude-in-chrome returning a screenshot is 1,855 tokens, while the median call returning only text is 373 tokens, with screenshot calls at about 5 times text calls.70% similarUnverifiedcheckwash 0.6.0 achieved a precision of 0.67 and recall of 0.22 for its own block judgment on this corpus, while the raw count signal read independently yielded a precision of 0.80 and recall of 0.89.68% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1096597API
curl https://kongchang.com/api/v1/knowledge/claims/1096597MCP
get_claim(id=1096597)