Zhipu AIGLM-4.5-Air
GLM-4.5-Air is a Zhipu AI open-weights reasoning model, released 20 Jul 2025. Its best configuration ranks #480 of 785 on the BenchLeader Index at 46.6 ±3.9, in the lower half. The ± is the point: 298 other configurations score within that range, so they and this one cannot be told apart on quality alone — price and speed are what separate them. It scores highest in human preference (54) and lowest in knowledge (32). At $0.373 per million tokens blended it is cheaper than most ranked models. Output speed of 77 tokens per second puts it slower than most, with a first answer in 28.6 s. Last measured 22 Oct 2026.
- Blended price
- $0.373/M
- $0.170 in · $0.980 out
- Output speed
- 77 tok/s
- measured by Artificial Analysis
- First answer
- 29 s
- first token 1.25 s
- Context
- 131k
- Full answer
- 35 s
- median, reasoning included
- Cost per run
- –
- Released
- 20 Jul 2025
- Overall index47
- Reasoning42
- Coding46
- Agents & tools52
- Maths53
- Knowledge32
- Instruction following46
- Human preference54
- Long context48
- Composite42
How its index has moved
8 Sept 2026 to 21 Oct 2026The index is recomputed from scratch every day, so a line moves when a new benchmark result lands, when a publisher revises a score, or when the models it is normalised against change. Early movement usually means the score is still settling.
Benchmark results
See all LLM benchmarks. One column per reasoning effort. Rank is among every configuration of every model on that benchmark. Hover a score for the run it came from.
Reasoning
| Benchmark | not stated | Source | Trend |
|---|---|---|---|
| Humanity's Last Exam | 8.1%#33 | Scale AI / CAIS | |
| Arena Hard Prompts | 1391#179 | Arena (LMArena) | |
| GPQA Diamond (AA)not in index | 73.3%#268 | Artificial Analysis | |
| Humanity's Last Exam (AA)not in index | 7.0%#373 | Artificial Analysis | |
| Kagi LLM Benchmark | 43.0%#99 | Kagi LLM Benchmark | |
| GPQA Diamond (HELM)not in index | 59.4%#28 | HELM Capabilities | |
| CritPt | 0.0%#315 | Artificial Analysis | |
| ForecastBenchnot in index | 62.0%#61 | ForecastBench |
Coding
| Benchmark | not stated | Source | Trend |
|---|---|---|---|
| GSO-Bench | 2.9%#32 | GSO-Bench | |
| Arena Coding | 1427#178 | Arena (LMArena) |
Agents & tools
| Benchmark | not stated | Source | Trend |
|---|---|---|---|
| Terminal-Bench Hard | 20.4%#163 | Artificial Analysis | |
| τ²-Bench Telecom (AA)not in index | 46.5%#205 | Artificial Analysis |
Maths
| Benchmark | not stated | Source | Trend |
|---|---|---|---|
| Omni-MATH (HELM) | 39.1%#31 | HELM Capabilities | |
| Arena Maths | 1387#167 | Arena (LMArena) |
Knowledge
| Benchmark | not stated | Source | Trend |
|---|---|---|---|
| AA-Omniscience | -61.5#474 | Artificial Analysis | |
| MultiNRCpossible contamination | 10.4%#41 | Scale AI SEAL | |
| MMLU-Pro (HELM)not in index | 76.2%#29 | HELM Capabilities | |
| AA-Omniscience: accuracynot in index | 16.3%#395 | Artificial Analysis | |
| AA-Omniscience: non-hallucinationnot in index | 7.1%#490 | Artificial Analysis |
Instruction following
| Benchmark | not stated | Source | Trend |
|---|---|---|---|
| IFBench | 37.5%#291 | Artificial Analysis | |
| IFEval (HELM) | 81.2%#42 | HELM Capabilities | |
| WildBench (HELM) | 78.9%#41 | HELM Capabilities | |
| Arena Instruction Followingnot in index | 1361#180 | Arena (LMArena) |
Human preference
| Benchmark | not stated | Source | Trend |
|---|---|---|---|
| Arena Text | 1373#181 | Arena (LMArena) | |
| Arena Creative Writingnot in index | 1327#189 | Arena (LMArena) | |
| Arena Multi-turnnot in index | 1367#181 | Arena (LMArena) | |
| Arena Longer Queriesnot in index | 1380#177 | Arena (LMArena) |
Long context
| Benchmark | not stated | Source | Trend |
|---|---|---|---|
| AA-LCR | 46.7%#331 | Artificial Analysis |
Composite
| Benchmark | not stated | Source | Trend |
|---|---|---|---|
| AA Intelligence Index v4.3.2 | 11.1#363 | Artificial Analysis | |
| HELM Capabilities meannot in index | 67.0%#30 | HELM Capabilities |
Safety & honesty
| Benchmark | not stated | Source | Trend |
|---|---|---|---|
| FORTRESSnot in index | 63.2%#2 | Scale AI SEAL | |
| MASKnot in indexpossible contamination | 60.8%#37 | Scale AI SEAL |
Where to run it
Every provider serving this model through OpenRouter, with throughput and first-token latency measured on live traffic over the last 30 minutes and each provider’s own price. Purple marks the best in each column.
| Provider | Speed | First token | Input $/M | Output $/M | Context | Quantisation |
|---|---|---|---|---|---|---|
| Z.ai | 71 tok/s | 1.42 s | $0.200 | $1.10 | 131k | fp8 |
| NovitaAI | 58 tok/s | 1.33 s | $0.130 | $0.850 | 131k | bf16 |
| SiliconFlow | 39 tok/s | 1.17 s | $0.140 | $0.860 | 131k | fp8 |
Price history
The price above, per 1M tokens, every day it changed.
What a task costs
Estimates from list price, output speed and time to first answer for the best configuration. “With caching” assumes three-quarters of the input is served from the prompt cache at $0.024 per 1M. Reasoning tokens are not modelled.
| Workload | Tokens in / out | Cost | With caching | Time |
|---|---|---|---|---|
| Chat reply | 400 / 300 | $0.0004 | $0.0003 | 32.5 s |
| Summarize a 30-page report | 12,000 / 600 | $0.0026 | $0.0013 | 36.4 s |
| Code edit | 6,000 / 1,500 | $0.0025 | $0.0018 | 48.2 s |
| Agentic coding session | 60,000 / 4,000 | $0.014 | $0.0075 | 1.3 min |
| Structured extraction | 2,000 / 200 | $0.0005 | $0.0003 | 31.2 s |
See also
Questions people ask
- What are GLM-4.5-Air's benchmark scores?
- As of 25 Oct 2026, GLM-4.5-Air scores 46.6 ±3.9 on the BenchLeader Index, strongest in human preference (54) and weakest in knowledge (32). That rests on 17 independent benchmark results, and every score on this page links to its publisher.
- How does GLM-4.5-Air rank?
- GLM-4.5-Air is #480 of 785 ranked configurations on 25 Oct 2026. 298 other configurations score within its margin, so the exact position matters less than the band. The leader is Claude Opus 5.5 (max effort) at 72.6.
- How much does GLM-4.5-Air cost per 1M tokens?
- As of 25 Oct 2026, GLM-4.5-Air costs $0.170 per 1M input tokens and $0.980 per 1M output tokens, $0.373 blended (three input tokens per output token). The list price last changed on 6 Oct 2026.
- When was GLM-4.5-Air released?
- Zhipu AI released GLM-4.5-Air on 20 Jul 2025.
Data as of 25 Oct 2026. Compare with another model.
Cite as: BenchLeader, “GLM-4.5-Air: benchmarks, pricing, speed and rank”, data as of 25 Oct 2026, https://www.benchleader.com/models/glm-4-5-air.