Zhipu GLM-5.2 crowned the first open-source AA Intelligent Index: GDPval benchmarking on par with GPT-5.5

robot
Abstract generation in progress

According to Beating Monitoring, Zhipu AI’s latest MoE flagship model GLM-5.2 scored 51 points in the Artificial Analysis Large Model Intelligence Index v4.1 evaluation, surpassing MiniMax-M3 (44 points), DeepSeek V4 Pro (max, 44 points), and Kimi K2.6 (43 points) to top the global open-source model rankings.

In the GDPval-AA v2 test simulating real-world knowledge work, GLM-5.2 scored 1524 points (human benchmark score 1000), leading MiniMax-M3 (1418 points) and DeepSeek V4 Pro (max, 1328 points), and tying with the closed-source frontier large model GPT-5.5 (xhigh reasoning). Compared with the previous-generation GLM-5.1, scientific reasoning CritPt increased by 16 percentage points to 21%, HLE rose by 12 percentage points to 40%, TerminalBench v2.1 improved by 16 percentage points to 78%, and GPQA Diamond reached 89%.

GLM-5.2 holds the best cost-effectiveness position on the “Intelligence - Task Cost” Pareto frontier. Since the average output per task is 43k tokens (26k for GLM-5.1), GLM-5.2’s average cost per task has increased to about 0.46 USD, higher than GLM-5.1 (0.25 USD) and DeepSeek V4 Pro (max, 0.05 USD), but still far lower than closed-source models in the same intelligence tier.

GLM-5.2 has 744B total parameters, 40B active parameters, and its context window has increased from 200K to 1M, following the MIT open-source license. Currently, Zhipu’s official API (pricing: input 1.4, output 4.4 / per million tokens) and platforms such as SiliconFlow, DeepInfra, and Nebius AI have already launched services.

View Original
This page may contain third-party content, which is provided for information purposes only (not representations/warranties) and should not be considered as an endorsement of its views by Gate, nor as financial or professional advice. See Disclaimer for details.
  • Reward
  • Comment
  • Repost
  • Share
Comment
Add a comment
Add a comment
No comments
  • Pinned