Zhipu AI has released GLM-5.3-Flash, the first natively multimodal model in the GLM-5 series, featuring visual coding capabilities and offering high performance at an extremely low price, setting a record for token usage on OpenRouter. The model scores 57 on the Artificial Analysis Intelligence Index, surpassing its predecessor and DeepSeek V4 Pro, while costing only one-fortieth of comparable models.
SynthePulse Insight · AI deep readingMembers
Version 1 · 1 source
Zhipu AI releases GLM-5.3-Flash, delivering flagship-level intelligence at 1/40 the price, and for the first time runs fully on domestic chips, opening a new chapter in AI competition.
The release of GLM-5.3-Flash breaks the law that 'larger parameters mean higher prices.' Its Artificial Analysis Intelligence score is 57, surpassing the previous flagship GLM-5.2 and DeepSeek V4 Pro's 53, but the price is only 0.8 yuan per million input tokens, 2.8 yuan per million output tokens, and 0.23 yuan for cache hits. In comparison, Claude Opus 4.8 costs about 35 yuan for input and 175 yuan for output per million tokens, making GLM-5.3-Flash roughly one-fortieth of its price.
Zhipu stated in the announcement: 'Same intelligence, 1/40 the price, frontier intelligence, no need to be frugal for the first time.' This pricing strategy is not merely marketing but is based on a lower cost structure, as Zhipu has validated the sustainability of low prices through engineering capabilities.
This article is based on a report from Titanium Media AGI, where model performance and pricing data come from official releases, but some details such as chip suppliers and quantities are not officially confirmed and are based on media reports.
Primary report
Primary source