GLM-5.3
Z.ai's open-weights flagship for complex coding and long-horizon tasks, with a 1M-token context window.
Start requirement: From 1 credit
Capabilities
Pricing
| Input | $2.8 / 1M tokens |
| Cached input | $0.52 / 1M tokens |
| Output | $8.8 / 1M tokens |
List prices plus 100%. Higher reasoning effort can consume more output tokens; it has no separate fixed multiplier. One credit equals 1¢ charged, with a one-credit minimum per completed run and the aggregate rounded up once.
Provider pricing source · verified 2026-09-10
GLM-5.3: frequently asked questions
- What is GLM-5.3?
- Z.ai's open-weights flagship for complex coding and long-horizon tasks, with a 1M-token context window. You can try GLM-5.3 on TryAI without a separate provider account.
- How much does GLM-5.3 cost on TryAI?
- GLM-5.3 uses these rates (provider list price plus 100%): Input: $2.8 / 1M tokens, Cached input: $0.52 / 1M tokens, Output: $8.8 / 1M tokens. One TryAI credit equals one cent charged; completed runs have a one-credit minimum and the aggregate cost is rounded up once.
- What can GLM-5.3 do?
- GLM-5.3 supports Streaming, Reasoning, Tool use with a context window of about 1M tokens.
- How do I try GLM-5.3?
- Create a free TryAI account, pick GLM-5.3 from the model selector, and start generating in seconds.
Related models
GPT-6 Astra
Chat & reasoning
250 credits to start
OpenAI's newest frontier model for complex reasoning, coding, and computer use, with a 1.05M-token context window. Self-reported 96.0 on GPQA and 91.5 on BrowseComp.
Try GPT-6 Astra→Claude Opus 5.5
Chat & reasoning
250 credits to start
Anthropic's newest Opus for long-running agentic coding and knowledge work. $4/$20 per million tokens, with cache reads at $0.20/M. Vendor-reported Terminal-Bench 4.0 66.4% and GDPval-AA 1846 Elo at max effort.
Try Claude Opus 5.5→Claude Fable 5.1
Chat & reasoning
250 credits to start
Anthropic's newest frontier model for coding, knowledge work, and long-horizon agentic runs. Beats both Fable 5 and Opus 5 on Terminal-Bench 4.0 (55.8%), CursorBench 3.2 (73.4%), and Humanity's Last Exam (65.0% with tools), at 75% cheaper cache reads than Fable 5.
Try Claude Fable 5.1→