EfficientLIVE PRICING

Z.ai: GLM 4.5 Air

text->text

Zhipu's flagship GLM-4. Strong on Chinese-language benchmarks, with frontier-class reasoning at competitive Chinese pricing.

Z.ai: GLM 4.5 Air is a efficient AI model from z-ai. It costs $0.130 per million input tokens and $0.850 per million output tokens (blended $0.346/M), with a 131,072-token context window.

Profile inherited from upstream GLM-4 Plus (4.5 variant) ↗ — this is a hosted variant of the same open-weights model.

INPUT
$0.130/M
per million input tokens
OUTPUT
$0.850/M
per million output tokens
BLENDED 70/30
$0.346/M
unchanged since 3 May
CONTEXT
131,072
tokens
What it is good at
  • Strong on Chinese benchmarks
  • Tool use
  • 128K context
  • Tightly integrated with Zhipu cloud
Typical use cases
  • Chinese-first production chat
  • Multilingual agents

Benchmarks

vs. best public score
Scores inherited from GLM-4 Plus (4.5 variant) — this is a hosted variant of the same open-weights model, so the underlying benchmark scores are identical.
MMLU84%
Multitask academic knowledge across 57 subjects.
Graduate-level science questions, "Google-proof".
MATH70%
High-school competition math problems.
Python function synthesis from docstrings.
LMArena Elo1244 Elo
Crowd-sourced head-to-head preference Elo rating.
Hand-curated from each provider's published reports and public leaderboards. Methodology varies across sources, treat as directional rather than authoritative.

More from z-ai

See all 6 →

Frequently asked questions

How much does Z.ai: GLM 4.5 Air cost?

Z.ai: GLM 4.5 Air costs $0.130 per million input tokens and $0.850 per million output tokens, for a blended reference rate of $0.346 per million tokens.

What is Z.ai: GLM 4.5 Air's context window?

Z.ai: GLM 4.5 Air supports up to 131,072 tokens of context in a single request.

What is Z.ai: GLM 4.5 Air best for?

Z.ai: GLM 4.5 Air is well suited to Strong on Chinese benchmarks, Tool use and 128K context.

Who makes Z.ai: GLM 4.5 Air?

Z.ai: GLM 4.5 Air is developed and served by z-ai.

Terms used on this page