glm-4.7-flash Benchmark & Insights

Zhipu AI Cloudflare Workers AI

Updated Jul 18, 2026 All models

Sample size

156 runs

in window

Accuracy

68.8%

consensus match · 160d

Confidence

97%

over 161 runs

Window end

Jul 18, 2026

most recent run

Input price

$0.06/MTok

prompt tokens

Output price

$0.40/MTok

completion tokens

Model insights

01 Cries wolf on roughly one day in five: 34 false "unsafe" calls and 28 "high" ratings against a consensus that saw only 9 "high" days.
02 It also managed the worst-case inversion, saying "safe" on the consensus-"unsafe" 2026-07-05.
03 Cheap, but the noise never quiets down — July was as bad as February.

Recent forecasts

Date

Conf.

Risk

Safe