GLM API Pricing
The Beijing lab behind the GLM models, also operating internationally as Z.ai. Spun out of Tsinghua University in 2019, Zhipu ships strong coding/agentic models, multiple free tiers, and open weights — and was the first of China's "Six Tigers" to pursue an IPO. It also sits on the US Entity List.
GLM-5.3
GLM-5.3 is priced as of 2026-08-23: $1.4/M input, $0.26/M cached input and $4.4/M output on Z.ai's International rate card, read in a rendered browser.
GLM-5.3-Flash
GLM-5.3-Flash is the first native multimodal model in Z.ai's GLM-5 series - 320B parameters with 18B activated, a 1M-token context, and a list price around a ninth of full-size GLM-5.3.
GLM-5.2 LAUNCHED JUN 2026
Z.AI pricing table lists GLM-5.2 at $1.40/M input, $0.26/M cached input, limited-time free cached input storage, and $4.40/M output.
GLM-5.1
Previous GLM flagship.
GLM-5
GLM-5 mid-tier flagship.
GLM-4 32B (0414, 128K)
Open-weight dated GLM-4 snapshot — 32B params, 128K context, symmetric $0.10/$0.10 pricing.
GLM-4.7
Cheaper sibling — same pricing as GLM-4.5/4.6.
GLM-4.7 FlashX
Ultra-cheap GLM-4.7 tier.
GLM-4.7 Flash
FREE tier for all registered Zhipu users.
GLM-4.6
Positioned by Zhipu as on-par with Claude Sonnet 4 on several benchmarks.
GLM-4.5 Air
Budget Zhipu tier.
GLM-4.5 Flash
FREE tier — Zhipu's older free model alongside GLM-4.7-Flash.
GLM-4.5 X
Premium GLM-4.5 variant — most expensive tier in the 4.5 family.
GLM-4.5 AirX
Mid-tier between GLM-4.5 Air ($0.20/$1.10) and GLM-4.5 X ($2.20/$8.90).
GLM-4.5 LAUNCHED JUL 2025
Z.AI pricing table lists GLM-4.5 at $0.60/M input, $0.11/M cached input, limited-time free cached input storage, and $2.20/M output.
GLM-5 Turbo
GLM-5 Turbo was the latency-optimised GLM-5 variant, priced between GLM-5 base ($1.00/$3.20) and GLM-5.1 ($1.40/$4.40) - and Z.ai does not sell it any more.
| Model | Input /M | Output /M | Cached | Context | Max output | Vision | Tools | Tier |
|---|---|---|---|---|---|---|---|---|
| GLM-5.3 FLAGSHIP | $1.40 | $4.40 | $0.26−81% | 1M | — | ✗ | ✓ | Frontier |
| GLM-5.3-Flash | $0.15 | $0.50 | $0.03−80% | 1M | — | ✓ | ✓ | Active |
| GLM-5.2 | $1.40 | $4.40 | $0.26−81% | 1M | — | ✗ | ✓ | Frontier |
| GLM-5.1 | $1.40 | $4.40 | $0.26−81% | 200K | — | ✗ | ✓ | Frontier |
| GLM-5 | $1.00 | $3.20 | $0.20−80% | 200K | — | ✗ | ✓ | Frontier |
| GLM-4 32B (0414, 128K) | $0.10 | $0.10 | — | 128K | — | ✗ | ✓ | Light |
| GLM-4.7 | $0.60 | $2.20 | $0.11−82% | 200K | — | ✗ | ✓ | Mid |
| GLM-4.7 FlashX | $0.07 | $0.40 | $0.01−86% | 200K | — | ✗ | ✓ | Light |
| GLM-4.7 Flash | Free | Free | Free | 200K | — | ✗ | ✓ | Light |
| GLM-4.6 | $0.60 | $2.20 | $0.11−82% | 200K | — | ✗ | ✓ | Mid |
| GLM-4.5 Air | $0.20 | $1.10 | $0.03−85% | 128K | — | ✗ | ✓ | Light |
| GLM-4.5 Flash | Free | Free | Free | 128K | — | ✗ | ✓ | Light |
| GLM-4.5 X | $2.20 | $8.90 | $0.45−80% | 128K | — | ✗ | ✓ | Frontier |
| GLM-4.5 AirX | $1.10 | $4.50 | $0.22−80% | 128K | — | ✗ | ✓ | Mid |
| GLM-4.5 | $0.60 | $2.20 | $0.11−82% | 128K | — | ✗ | ✓ | Active |
| GLM-5 Turbo DEPRECATED | $1.20 | $4.00 | $0.24−80% | 200K | — | ✗ | ✓ | Deprecated |
Pricing across the lineup.
The GLM story so far.
Releases and corporate moves from Zhipu / Z.ai. Sourced from Z.ai docs, Zhipu press, and Wikipedia — verified at publication.
docs.z.ai — the international Z.ai brand and API for the full GLM catalog, including free Flash tiers. OpenAI-compatible; cache-hit input is a fraction of base.
CHINA · DOMESTICopen.bigmodel.cn — Zhipu's China-domestic platform (BigModel / 智谱), keeping data inside mainland China for domestic compliance.
OPEN-WEIGHT · SELF-HOSTGLM open weights — GLM-4.5, GLM-4.6, and the GLM-4 32B snapshot are published for self-hosting, fully outside any China-routing of your data.
Zhipu AI (智谱) was spun out of Tsinghua University in 2019 by professors Tang Jie and Li Juanzi, and is led by CEO Zhang Peng. It develops the GLM (General Language Model) series with Tsinghua's KEG lab, and operates internationally under the Z.ai brand. It is one of China's "Six Little Tigers" of AI.
Zhipu's lineup is unusually broad and price-aggressive: the GLM-5.3 flagship at $1.40/$4.40 with its native-multimodal sibling GLM-5.3-Flash a ninth of that, the earlier GLM-5.2 and GLM-5.1 at the same flagship rate, GLM-5 base below them, the coding-focused GLM-4.7 (compatible with Claude Code, Cline, Roo Code), a premium GLM-4.5 X tier, and — distinctively — two free models (GLM-4.7 Flash and GLM-4.5 Flash) plus an ultra-cheap GLM-4.7 FlashX at $0.07/M. Several models, including GLM-4.5 and GLM-4.6, are released as open weights.
On funding, Zhipu raised from Alibaba, Tencent, Ant Group, and Saudi Arabia's Prosperity7, with a valuation pushed above $20 billion by 2025. It was the first of the "Six Tigers" to complete an IPO, listing in early 2026.
Two facts must be stated plainly. First, in January 2025 the US Commerce Department added Zhipu to the Entity List over national-security concerns — which restricts its access to US technology and matters for some Western buyers' procurement. Second, in February 2026 Zhipu said it had trained a major model on Huawei chips, a response to those very chip restrictions.
The trade-offs: most of the priced lineup here is text/code-focused — vision arrived at the top of the range with GLM-5.3-Flash, and Z.ai's separate GLM-4.6V family is not yet in this catalogue — the direct China platform stores data under Chinese rules, and the Entity-List status is a real procurement consideration. But Zhipu's open weights and free tiers let you sidestep the hosted API entirely. Versus DeepSeek and Moonshot, Zhipu competes on coding strength, free access, and breadth.
DeepSeek
MIT open weights at the lowest cost-per-token. Comparable open-weight stance; Zhipu counters with free tiers and stronger agentic-coding tooling integration.
CHINESE PEERAlibaba (Qwen)
Broad open-weight catalog with 119-language coverage and multimodality. Qwen is wider and multimodal; Zhipu is leaner and coding-focused with free tiers.
SIX-TIGERS PEERMoonshot (Kimi)
1T-param open-weight MoE and long-context heritage. A fellow Beijing "Tiger"; Moonshot chases scale, Zhipu chases breadth + free access.
PRIMARY RIVALOpenAI
Frontier benchmarks + global ecosystem (GPT-5). Closed and far pricier; Zhipu's GLM-4.6 is pitched as Claude-Sonnet-class on coding at a fraction of the cost.
FRONTIER LEADERAnthropic
Leads coding + reasoning with Claude. Notably, GLM models are tuned to plug into Claude Code — Zhipu targets the same agentic-coding niche far cheaper.
CHINA CONSUMER LEADERByteDance (Doubao)
The Doubao model family from ByteDance — parent of TikTok and Douyin. Served through the Volcano Ark (火山方舟) platform on ByteDance's Volcano Engine cloud, Doubao powers China's most-used consumer AI app and undercuts most frontier labs on price.
Frequently asked.
Practical questions about GLM pricing, free tiers, open weights, and sanctions.
Q · 01 Which GLM model should I start with? +
$0.60/$2.20) — it plugs into Claude Code, Cline, and similar tools. For the frontier: GLM-5.2 ($1.4/$4.4). Prototyping for free? GLM-4.7 Flash is $0/M. See the use case picker above.Q · 02 Are there really free models? +
$0/M) for registered users, rate-limited but with no per-token charge and up to 200K context. The cheapest paid tier is GLM-4.7 FlashX at $0.07/$0.40. Few labs offer free frontier-adjacent tiers like this.Q · 03 Are GLM models open-weight? +
Q · 04 Is Zhipu under US sanctions? +
Q · 05 Where is my data stored? +
Q · 06 Why is GLM popular for coding? +
$0.60/$2.20, GLM-4.7 is a popular low-cost backend for coding agents.