Z AI has released GLM-5-Turbo, a proprietary model optimized for agentic use cases that scores lower than GLM-5 (Reasoning) on the Artificial Analysis Intelligence Index

@Zai_org's GLM-5-Turbo scores 47 on the Artificial Analysis Intelligence Index, 3 points behind the open weights GLM-5 (Reasoning, 50). GLM-5-Turbo scores 1503 on GDPval-AA, ahead of GLM-5 (Reasoning, 1408). GLM-5-Turbo is a text-only proprietary model with a 200K token context window, available via the Z AI first-party API. Z AI claims GLM-5-Turbo is "a foundation model deeply optimized for the OpenClaw scenario" and it scores higher than GLM-5 on GDPval-AA

Key takeaways:

➤ GLM-5-Turbo scores 47 on the Intelligence Index, 3 points behind GLM-5 (Reasoning, 50). The gap is driven by weaker performance on TerminalBench (33% vs 43%), CritPt (0.3% vs 2%), and HLE (25% vs 27%). GLM-5-Turbo scores marginally higher on GPQA (85% vs 82%) and IFBench (73% vs 72%). This places GLM-5-Turbo equivalent in intelligence to Kimi K2.5 (Reasoning, 47)

➤ The model scores 1503 on GDPval-AA (Agentic Real-World Work Tasks) ahead of GLM-5 (Reasoning, 1408), MiniMax-M2.7 (1495), and GPT-5.4 mini (xhigh, 1407). This is directionally in line Z AI’s claims that the model has been specifically optimized for agentic task performance

➤ GLM-5-Turbo scores lower on AA-Omniscience compared to GLM-5: GLM-5-Turbo scores -15.1 vs GLM-5 (Reasoning, +2.0), a lower score in knowledge reliability and hallucination

➤ GLM-5-Turbo used ~94M output tokens to run the Intelligence Index, compared to ~109M for GLM-5 (Reasoning), making it slightly more token-efficient. However, GLM-5-Turbo is priced higher at $1.20/$4.00 per 1M input/output tokens vs GLM-5 at $1.00/$3.20, resulting in a slightly higher effective cost to run (~$584 vs ~$547) despite the lower token usage

➤ Both GLM-5-Turbo and GLM-5 are text-input and output only. Unlike the open weights, MIT-licensed GLM-5, GLM-5-Turbo is proprietary and not available for self-hosting. Both models share a 200K token context window