GLM-5.3 Smart Index Skyrockets to 60: Ties Kimi K3, API Price Unchanged

According to Watchful AI Monitoring, MindNet has officially launched the GLM-5.3 API. The API was not live immediately upon the model's release. At the same time, the Independent Evaluation from Artificial Analysis has been released: the Intelligence Index surged from 53 in GLM-5.2 to 60, tying with Kimi K3 and only 1 point below GPT-5.6 Sol. Once the equal weights are publicly disclosed, it will tie with Kimi K3 as the highest-rated open-weight model in the AA ranking.

The biggest gainer was Agent. The Elo score of GDPval-AA v2 jumped from 1524 to 1770, a 246-point increase in one go, ranking second among all models, just behind Claude Opus 5 with 1855 and over 100 points higher than Kimi K3 at 1668.

The API pricing did not increase accordingly. The input remains at $1.40 per million Tokens, output at $4.40, and cached input at $0.26, all the same as GLM-5.2.

However, 5.3 clearly requires more Tokens. AA found that it outputs an average of about 18,700 Tokens per task, 20% more than 5.2. Therefore, the actual cost per task has increased from $0.44 to $0.68, a price hike of around 55%. Even so, it is still cheaper than Kimi K3's $0.84 and GPT-5.6 Sol's $1.23 among models of the same tier.