← All Models

Zhipu AI

GLM-5

Zhipu AI's 744B MoE flagship. Open-weight under MIT license. Strong on SWE-Bench and agentic tasks.

flagship

Input price

$1/1M

Output price

$3.2/1M

Context window

200K

Max output

128K

Capabilities

textvisiontool-usecodereasoning

Benchmarks

MMLU-Pro70.4%
HumanEval91%
GPQA72%

Strengths

  • Open-weight (MIT license) — self-hostable
  • 77.8% SWE-Bench Verified — top-tier coding
  • 128K max output — huge generation window

Weaknesses

  • MMLU-Pro lags behind Western flagships
  • 744B parameters — heavy to self-host
  • China-based — availability concerns

Compare with similar models