Qoder Thinking Effort
3 pasos1 piezas1 Fuentes
Alibaba Cloud's Qoder platform has adjusted the default "thinking effort" for four AI models—GLM-5.3, Kimi-K3 (now replaced by Kimi-K3.1-Pro), Kimi-K2.8-Preview, and DeepSeek-Flash—from "Maximum" to "High". This change, based on internal benchmarks using QoderBench, aims to balance performance with cost, reportedly reducing costs by up to 40.5% for these specific tasks. The move highlights a growing industry tension: while higher reasoning depth can improve accuracy on complex problems, it often yields diminishing returns on simpler tasks while significantly increasing latency and credit consumption. Users retain the ability to manually revert these settings via the model selector. The core disagreement lies in whether automated defaults should prioritize speed/cost or maximum potential accuracy, a trade-off that varies heavily by use case.
Cómo llegó hasta aquí
Alibaba Cloud's Qoder team evaluated six models on real tasks using QoderBench, comparing 'Maximum' vs 'High' thinking effort levels.
alibabacloud.com ↗Based on results showing diminishing returns for many tasks, Qoder changed the default thinking effort for GLM-5.3, Kimi-K3, Kimi-K2.8-Preview, and DeepSeek-Flash from 'Maximum' to 'High'.
alibabacloud.com ↗Kimi-K3 was subsequently replaced by Kimi-K3.1-Pro in the model list, which inherits the 'High' default setting pending further evaluation.