Quick Take

GLM-5.2 (max), released by Z AI on June 16, 2026, and Google’s Gemini 3.5 Flash (high), released May 19, 2026, are both high-performance models. GLM-5.2 (max) holds a slight edge in general intelligence (50.7 vs 50.2) and coding (50.7 vs 45). Gemini 3.5 Flash (high) excels in raw output speed, though it faces a significantly higher time to first token.

Benchmark Read

GLM-5.2 (max) demonstrates strong capabilities in technical tasks, scoring 0.507 on TerminalBench Hard and 0.991 on TAU2. Gemini 3.5 Flash (high) outperforms its rival in several key areas, including GPQA (0.922 vs 0.895), HLE (0.41 vs 0.401), SciCode (0.531 vs 0.505), and IFBench (0.763 vs 0.733). However, GLM-5.2 (max) remains more effective in LCR (0.713 vs 0.693) and TerminalBench Hard (0.507 vs 0.409).

Cost and Speed

Cost structures differ significantly between the two providers. GLM-5.2 (max) is more affordable, with a blended price of $2.15/1M tokens, compared to Gemini 3.5 Flash (high) at $3.38/1M tokens.

Regarding performance, Gemini 3.5 Flash (high) delivers a much faster output speed of 205.086 tok/s compared to GLM-5.2 (max)’s 114.156 tok/s. Conversely, GLM-5.2 (max) is significantly more responsive, with a time to first token of 2.085s, whereas Gemini 3.5 Flash (high) requires 16.451s to begin generating output.

Best Fit

GLM-5.2 (max) is best suited for developers requiring high coding precision and low-latency responses. Gemini 3.5 Flash (high) is better suited for high-throughput applications where the speed of long-form content generation outweighs the initial delay in response time.