Unverified50% confidenceBenchmarkExact time
According to benchmark data released by OpenRouter, the Jeff model takes only 154 milliseconds to make a decision, while the second-place model requires 860 milliseconds.
1
Sources
50%
Confidence
Medium-term (~90 days)
Relevance
9/22/2026
First Seen
Valid until: 12/21/2026
Sources
Related Entities
Related Claims
Unverified参数量在1亿以下的轻量模型经领域数据微调后可达95%以上意图分类准确率，单次推理延迟控制在5-20ms以内65% similarUnverified针对有限意图集合训练的轻量级专用模型参数量通常在10M以下，其推理延迟可比通用LLM调用低2-3个数量级63% similarUnverifiedAccording to official performance tests, a search on a graph with 118,000 nodes saw response time drop from 24 milliseconds to 0.004 milliseconds, a speedup of approximately 6,000 times.63% similarUnverifiedGLM 5.2's inference speed is approximately 106 tokens per second.60% similarUnverifiedTiny-MoE模型总参数量约200M，但每个token激活的参数仅约90M60% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/941068API
curl https://kongchang.com/api/v1/knowledge/claims/941068MCP
get_claim(id=941068)