Skip to content
llmwaves

GLM 4.5 Air

Z.ai · released Jul 25, 2025

Open weightsReasoningTool use

GLM-4.5-Air is the lightweight variant of our latest flagship model family, also purpose-built for agent-centric applications. Like GLM-4.5, it adopts the Mixture-of-Experts (MoE) architecture but with a more compact parameter...

Specification

Context window
131K
Max output
98K
Knowledge cutoff
Dec 31, 2024
Parameters
110.5B
Licence
mit
Serving providers
3
Moderated
No
Uptime
100.0%

Intelligence

16.7

52th percentile

Coding

Coding Index

Agentic

Agentic Index

Output speed

42 t/s

Median across providers

Latency

782ms

Time to first token

Cost per task

$0.07

Estimated

Benchmarks

Where the score comes from

The Intelligence Index is a composite. These are the underlying evaluations this model was actually measured on.

Evaluation scores

Percentage correct · higher is better

  • MMLU-Pro
    81.5%
  • AIME 2025
    80.7%
  • GPQA Diamond
    73.3%
  • LiveCodeBench
    68.4%
  • τ²-bench (Telecom)
    46.5%
  • AA-LCR (long context)
    45.7%
  • IFBench
    37.6%
  • SciCode
    30.6%
  • Terminal-Bench Hard
    20.5%
  • Humanity's Last Exam
    7.0%

An evaluation missing from this list was not run for this model — it is not a zero.

View as table
EvaluationScore
MMLU-Pro81.5%
AIME 202580.7%
GPQA Diamond73.3%
LiveCodeBench68.4%
τ²-bench (Telecom)46.5%
AA-LCR (long context)45.7%
IFBench37.6%
SciCode30.6%
Terminal-Bench Hard20.5%
Humanity's Last Exam7.0%

Against its peers

Intelligence Index · this model highlighted, nearest peers in grey

Peers are the models sitting closest on the Intelligence Index — the set you would realistically choose between.

View as table
ModelIntelligence
Trinity Large Thinking18.6
Sonar Reasoning Pro18.0
GLM 4.5 Air16.7
o3 Mini High15.7
gpt-oss-20b15.2
gpt-oss-20b (free)15.2
Nemotron 3 Nano Omni (free)15.0
GPT-4.1 Mini14.8

Percentile among all indexed models

Intelligence52th
Mathematics77th
Terminal work60th
Arena Elo63th

Pricing

What it costs to run

List prices per million tokens, plus what one representative task works out to.

List price

Input / 1M tokens
$0.13
Output / 1M tokens
$0.85
Cached input / 1M
$0.025
Blended 3:1
$0.31

One task, estimated

$0.07

Input tokens
50,000
Output tokens
80,000
Profile
Reasoning

Estimated from list pricing: 50K input tokens plus 80K output tokens for reasoning models (25K for non-reasoning).

Serving providers

Speed and latency figures are medians across these providers, so a widely-served model reports a blend rather than any single endpoint.

Arena

Head-to-head generation quality

Elo from pairwise judgements, broken out by the kind of thing the model was asked to build.

Overall Elo
1,218
Win rate
59.2%
Strongest at
Data visualisation
Tournaments
218

Elo by category

Dot position on a 1,100–1,225 scale · Elo has no meaningful zero

Agent categories (full-stack apps, mobile apps) are only scored for models tested in agent harnesses.

View as table
CategoryEloWin rate
Data visualisation121859.2%
3D scenes117854.1%
Websites116951.3%
UI components115854.5%
Game development113548.4%
SVG111850.8%