Unverified50% confidenceBenchmarkExact time
Laya is an open-source small model with 421 million parameters that can achieve a score of 60 points against single-digit results within 30 seconds on simple judgment tasks of the greedy-tongue type.
1
Sources
50%
Confidence
Medium-term (~90 days)
Relevance
9/22/2026
First Seen
Valid until: 12/21/2026
Sources
Related Entities
Related Claims
UnverifiedThe algorithm problems used in the evaluation came from Huawei's coding platform, with a difficulty level of approximately 200 points, classified as medium-to-high difficulty.67% similarUnverified根据文中量化估算，编程智能体任务中模型推理仅占总时间的 20%-30%，剩余 70%-80% 花在等待外部系统 I/O 响应66% similarUnverifiedFor tiny models with only a few layers that complete in a few milliseconds, the framework overhead of a general-purpose inference engine can account for more than 50% of the total time.66% similarUnverified阿里千问新一代模型参数达2.4万亿，是该系列首个突破万亿的多模态大模型，代码生成准确率据称达到94.7%，较上一代提升约11个百分点65% similarUnverified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% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/941069API
curl https://kongchang.com/api/v1/knowledge/claims/941069MCP
get_claim(id=941069)