- Published on
GLM-5.3 hits the API at $1.4/$4.4 per million tokens
- Authors
- Name
- DeepHorus
Chinese startup z.ai has made its frontier open-source language model GLM-5.3 available through its API, following a debut last week that showcased advanced cyber capabilities. Developers can now integrate the model into their agents and applications, though current GLM Coding Plan subscribers are limited to the OpenAI Chat Completions-compatible protocol for now. The company says open-weight release is planned, but no date or license has been confirmed.
Pricing remains unchanged from the previous GLM-5.2 generation: $1.40 per million input tokens and $4.40 per million output tokens. Cached input costs $0.26 per million tokens, and storage is free for a limited time. This means developers can upgrade to the new model without facing a higher headline rate, even as z.ai claims substantially stronger coding and long-horizon agent performance.
A simple comparison of one million input and one million output tokens places GLM-5.3 at $5.80, well below rivals such as Grok 4.6 at $8, Kimi K3 at $18, Claude Opus 5 at $30, and GPT-5.6 Sol at $35. However, that is not a workload-cost estimate, as real bills depend on input/output mix, caching, and token consumption. Still, the price puts GLM-5.3 in a notably lower cost band than premium frontier models.
Independent results highlight the model's competitiveness. Artificial Analysis gives GLM-5.3 a score of 60 on its Intelligence Index, tying Kimi K3 as the top performing open-weights model globally and scoring seven points higher than GLM-5.2. Yet its analysis estimates GLM-5.3 costs about $0.68 per Intelligence Index task versus roughly $0.44 for GLM-5.2, despite identical token prices.
That difference underscores a key caveat: GLM-5.3 is more verbose than its predecessor, so flat per-token rates do not guarantee flat costs for completed workloads. For developers, the immediate takeaway is straightforward—GLM-5.3 is now callable through z.ai's API at the same $1.40/$4.40 rate, offering another relatively low-cost option for frontier-class coding and agent tasks. The move signals z.ai's intent to compete aggressively on price while pushing open-weight performance forward.
