GLM-4.5-Air: benchmark scores, price and ranking

GLM-4.5-Air from Z.ai (Zhipu AI),Tsinghua University ranks #327 of 743 LLMs on the LLMs Tiger aggregate benchmark score (48.7), based on 4 benchmark results captured up to 2026-10-13.

Organisation: Z.ai (Zhipu AI),Tsinghua University
Type: Open-weight / local model
Open weights: Yes
Released: 2025-07-20 (source: epoch)
Cheapest API price: $0.125 in / $0.450 out per million tokens (via pinstripes, observed 2026-10-13T20:47Z)
Aggregate rank: #327 of 743 (score 48.7)

Relative to other models, GLM-4.5-Air is strongest on MMLU-Pro (#29 of 136) and weakest on GSO-Bench (#16 of 19).

Across the 4 benchmarks where at least ten models are scored, GLM-4.5-Air is in the top 10% on 0 and in the top half on 1.

Its cheapest listed input price, $0.125 per million tokens, is lower than that of 77% of the 427 models with a listed price.

GLM-4.5-Air benchmark results

GLM-4.5-Air benchmark scores, ranked against all models measured on each benchmark
Benchmark	Score	Rank	Source	Captured
MMLU-Pro	81.4%	#29 of 136	HuggingFace MMLU-Pro leaderboard	2026-10-13
GPQA Diamond	71.7%	#138 of 275	HuggingFace GPQA Diamond leaderboard	2026-10-13
Humanity's Last Exam	8.1%	#16 of 27	Epoch AI Benchmarking Hub	2026-10-13
GSO-Bench	2.9%	#16 of 19	Epoch AI Benchmarking Hub	2026-10-13

About GLM-4.5-Air

GLM-4.5-Air is an open-weight model that can run on your own hardware from Z.ai (Zhipu AI),Tsinghua University. It was released on 2025-07-20, according to Epoch AI. The cheapest API listing we track charges $0.125 per million input tokens and $0.450 per million output tokens, through pinstripes. We hold 4 benchmark results for it across 2 areas, and none is in the top tenth of its board.

Where GLM-4.5-Air stands

On our aggregate score, GLM-4.5-Air is #327 of 743 ranked models, around the middle of the table. Its score is 48.7; the scale is centred on 50, so it sits 1.3 points below the typical model. That beats 56% of the ranked field. Among the 4 ranked Z.ai (Zhipu AI),Tsinghua University models it is #2. The next one up is GLM-4.5 Thinking at 49.2. The next one down is glm-4.5 at 48.4.

GLM-4.5-Air, benchmark by benchmark

MMLU-Pro puts GLM-4.5-Air at 81.4%, #29 of 136. That beats 79% of the models tested, but 6.7 points short of Darwin-180B-RSI (88.1%). The results either side are A.X-K1 (81.5%) and DeepSeek V3 0324 (81.2%). It covers 57-subject knowledge + reasoning. The board median is 55.3%, so GLM-4.5-Air is 26.1 points above it. Source: HuggingFace MMLU-Pro leaderboard, captured 2026-10-13.

GLM-4.5-Air lands below the middle on GPQA Diamond: 71.7%, #138 of 275, with 50% of the field scoring higher. GPT-6 Astra (max) leads at 95.8%; the closest result above is qwen3-30b-a3b-thinking-2507 at 72.2%. It covers expert-level science questions. The median score on this board is also 71.7%. Source: HuggingFace GPQA Diamond leaderboard, captured 2026-10-13.

GLM-4.5-Air lands below the middle on Humanity's Last Exam: 8.1%, #16 of 27, with 56% of the field scoring higher. Muse Spark leads at 40.6%; the closest result above is glm-4.5 at 8.3%. It covers expert-level questions across domains. The board median is 10.7%, so GLM-4.5-Air is 2.6 points below it. Source: Epoch AI Benchmarking Hub, captured 2026-10-13.

GSO-Bench is a weak spot. GLM-4.5-Air scores 2.9%, #16 of 19, behind 79% of the models measured. The leader, GPT-5.5 (xhigh), reaches 40.2%. It covers software performance optimisation. The board median is 4.9%, so GLM-4.5-Air is 2.0 points below it. Source: Epoch AI Benchmarking Hub, captured 2026-10-13.

Strengths and weaknesses

Reasoning, maths and science: solid, ahead of 56% of the field on average across 3 benchmarks. Its best board here is MMLU-Pro (#29 of 136) and its weakest is Humanity's Last Exam (#16 of 27).

Price and value

The median input price across the 427 priced models is $0.500 per million tokens, so GLM-4.5-Air costs less than the typical listing. Models that score within 1.5 points or better and cost less: Qwen2.5-Coder-7B-Instruct ($0.010, 49.5), DeepSeek V4.1 Flash (high) ($0.010, 48.9), DeepSeek V4 Flash 0731 (max) ($0.016, 53.1). Similar scores cost more on Qwen2-VL-72B-Instruct ($0.130) and Hunyuan-A13B-Instruct ($0.140).

Release history

The next one was glm-4.5 on 2025-07-28, at 48.4.

Reading these numbers

The 4 scores on this page come from 3 sources: Epoch AI Benchmarking Hub (2), HuggingFace GPQA Diamond leaderboard (1), HuggingFace MMLU-Pro leaderboard (1). No official leaderboard supplied any of them, so treat them as third-party measurements. With 4 results the aggregate rank can move a lot when one new score arrives.

Models ranked nearby

#324 Opus 4.5 (Thinking, 16K) (48.8)
#325 gemini-2.5-flash-lite-preview-06-17_16K (48.7)
#326 Qwen3-30B-A3B-Base (48.7)
#328 GPT-5.1 Codex Mini (48.7)
#329 Amazon.nova Premier v1:0 (48.7)
#330 MiniMax-M2 (48.7)