NOUMENOpen app
NOUMEN

Z.ai · Open weights

GLM 4.5 Air

Added Apr 15, 2025Open weights

Compare models

Choose up to 3 models to compare.1 of 3 chosen

Loading models…

Use in the app

GLM-4.5-Air is a 106B total / 12B active parameter model designed to unify frontier reasoning, coding, and agentic capabilities. On the SWE-bench Verified benchmark, it delivers the best performance at its scale with a competitive performance-to-cost ratio.

  • Tools
  • Structured output
Input
Text
Output
Text
Discovery tier
Open weights

Pricing

USD per 1M tokens
Input
$0.13
Output
$0.85
Cache read
$0.025

Reference prices as published by the model provider. API example

Specifications

Context window
128K
Max output
98.3K
Average output (7 days)
59 tokens
Longer than 5% of 465 models

Side by side

GLM 4.5 Air compared

Specifications, prices, speed and benchmark scores next to each other. Benchmarks below show every compared model too.

Compare models

Choose up to 3 models to compare.1 of 3 chosen

Loading models…

GLM 4.5 AirZ.aiThis modelGranite 4.2 8BibmRemove
ProviderZ.aiibm
Discovery tierOpen weightsOpen weights
AddedApr 15, 2025Aug 31, 2026
Limits
Context window128K131.07KBest
Max output98.3K—
Average output59 tokens2.4K tokens
Pricing, USD per 1M tokens
Input$0.13$0.10Best
Output$0.85$0.15Best
Cache read$0.025Best$0.05
Performance
Throughput89.9 tok/s—
Time to first token1.6 s—
Request success (24h)100.0%100.0%
Benchmark indexes
Intelligence Index11.111.1
Coding Index—22.4
Math Index80.7—
Capabilities
ReasoningNoYes
ToolsYesYes
Structured outputYesYes
Reasoning effort—none, low, high

Public evaluations

Benchmarks

Independent evaluation results, with how each score ranks among the models in the Noumen catalog.

  • GLM 4.5 Air
  • Granite 4.2 8B

Intelligence Index

  • GLM 4.5 Air11.1
  • Granite 4.2 8B11.1

GLM 4.5 Air: Better than 23% of 326 scored models

3.8Catalog models by score55.1

Indexes

  • Coding Index

    Composite of the coding evaluations

    • GLM 4.5 Air: —
    • Granite 4.2 8B: 22.4

  • Math Index

    Composite of the math evaluations

    • GLM 4.5 Air: 80.7
    • Granite 4.2 8B: —

Agentic work

  • AutomationBench-AA

    Workflow automation, penalized for guardrail violations

    • GLM 4.5 Air: —
    • Granite 4.2 8B: 2.1%

  • AutomationBench-AA tasks completed

    Workflows finished end to end without a guardrail violation

    • GLM 4.5 Air: —
    • Granite 4.2 8B: 0.3%

  • AA-Briefcasev1.1

    Agentic knowledge work, rated head to head

    • GLM 4.5 Air: —
    • Granite 4.2 8B: 310 Elo

  • GDPval-AAv2.1

    Economically valuable occupational tasks, rated head to head

    • GLM 4.5 Air: —
    • Granite 4.2 8B: 491 Elo

Document reasoning

  • GDP.pdf

    Professional reasoning over PDF documents, all-pass rate

    • GLM 4.5 Air: —
    • Granite 4.2 8B: 0.6%

  • AA-LCRv1.1

    Reasoning across long documents

    • GLM 4.5 Air: 46.7%Top
    • Granite 4.2 8B: 45.0%

Reasoning and knowledge

  • Humanity’s Last Exam

    Expert-level questions across academic fields

    • GLM 4.5 Air: 7.0%
    • Granite 4.2 8B: 9.7%Top

  • GPQA Diamond

    Graduate-level science questions

    • GLM 4.5 Air: 73.3%Top
    • Granite 4.2 8B: 63.1%

  • MMLU-Pro

    Multi-subject knowledge and reasoning

    • GLM 4.5 Air: 81.5%
    • Granite 4.2 8B: —

  • IFBench

    Following precise output instructions

    • GLM 4.5 Air: 37.6%
    • Granite 4.2 8B: —

Coding

  • Terminal-Benchv4.0

    Practical software engineering in a terminal

    • GLM 4.5 Air: —
    • Granite 4.2 8B: 0.0%

  • SciCode

    Scientific computing in Python

    • GLM 4.5 Air: —
    • Granite 4.2 8B: 31.5%

  • LiveCodeBench

    Recent competitive programming problems

    • GLM 4.5 Air: 68.4%
    • Granite 4.2 8B: —

Math

  • AIME 2025

    Competition mathematics, 2025 problems

    • GLM 4.5 Air: 80.7%
    • Granite 4.2 8B: —

  • AIME

    Competition mathematics

    • GLM 4.5 Air: 67.3%
    • Granite 4.2 8B: —

  • MATH-500

    Competition math problems

    • GLM 4.5 Air: 96.5%
    • Granite 4.2 8B: —

Legacy evaluations

  • Terminal-Bench Hard

    Earlier terminal task suite

    • GLM 4.5 Air: 20.5%
    • Granite 4.2 8B: —

  • τ²-Bench Telecom

    Tool use in a customer-support simulation

    • GLM 4.5 Air: 46.5%
    • Granite 4.2 8B: —

  • AA-LCR (unversioned)

    Long-context reasoning, earlier grading

    • GLM 4.5 Air: 46.7%Top
    • Granite 4.2 8B: 45.0%

Scores from Artificial Analysis, updated Oct 26, 2026. Artificial Analysis

Live routing data

Performance and reliability

Recent averages across requests routed for this model. Your own speed depends on prompt length, reasoning effort and load.

Throughput
89.9 tok/s
Faster than 57% of 251 models
Time to first token
1.6 s
Quicker than 85% of 251 models
Prompt cache hits
40.6%
About $0.0873 per 1M input tokens with caching
Request success (24h)
100.0%
Reliable

Cache hits are the share of input tokens recently served from a prompt cache across all requests. They are not a forecast of your own hit rate.

Hourly request success · last 72 hoursLowest hour 100.0%
  • Healthy (98% or more)
  • Degraded (90–98%)
  • Down (under 90%)
  • No requests

Developers

Use it through the API

Noumen’s API is OpenAI-compatible: any Chat Completions client works with a base URL and a key.

Read the quickstart
curl https://www.noumen.xyz/api/v1/chat/completions \
  -H "Authorization: Bearer $NOUMEN_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "z-ai/GLM-4.5-Air",
    "messages": [{ "role": "user", "content": "Explain this stack trace." }]
  }'
import os
from openai import OpenAI

client = OpenAI(base_url="https://www.noumen.xyz/api/v1", api_key=os.environ["NOUMEN_API_KEY"])
reply = client.chat.completions.create(
    model="z-ai/GLM-4.5-Air",
    messages=[{"role": "user", "content": "Explain this stack trace."}],
)
print(reply.choices[0].message.content)

Z.ai

Other Z.ai models, starting with the same family.