GLM-5.3 API: pricing, base URL and quick setup
Use GLM-5.3 on A2Agent with model ID glm-5.3. The catalog lists a 1M context window. Check token rates below and verify the features your app needs.
Specifications and pricing
| Model ID | glm-5.3 |
|---|---|
| Platform | Z.ai |
| Context window | 1M |
| Tagged for | chat, coding |
| Discount | -40% |
| Input / 1M tokens | |
| Output / 1M tokens |
USD per 1M tokens. Final billing depends on your API key group and metered usage.
What GLM-5.3 costs on A2Agent
The official price is $1.40 input and $4.40 output per 1M tokens. On A2Agent it is $0.840 and $2.64. A month of 10M input and 2M output tokens costs $22.80 at the official price and $13.68 on A2Agent.
Good to know
- The reference rates in the table match the Z.ai international price list as of 2026-11-06.
Where GLM-5.3 fits
Long-running coding agents
Z.ai positions GLM-5.3 for complex software engineering and agent tasks. It fits multi-file changes and terminal-driven work in Claude Code or Codex CLI, where the model plans, edits and retries.
Basis: Z.ai model page; Terminal Bench 3.0 and DeepSWE results below.
Large repositories and long documents
The catalog lists a 1M context window and Z.ai states a 128K maximum output. A whole module, its tests and a long issue thread can sit in one request.
Basis: Catalog context field; Z.ai model page for the output limit.
Not the pick for quick, low-cost replies
Z.ai states that reasoning is always on for GLM-5.3 and cannot be switched off. Short lookups and classification still spend reasoning tokens. GLM-5.3 Flash is listed at a much lower rate for that work.
Basis: Z.ai model page; compare rates in the table above.
Vendor-reported benchmarks
These figures come from the vendor, not from A2Agent tests. Harnesses, reasoning effort and sampling differ between vendors, so do not rank models by them. Checked 2026-11-06.
| Benchmark | Score | Vendor test conditions |
|---|---|---|
| Terminal Bench 3.0 | 28.3 | Claude Code 2.1.207 harness, max reasoning effort, 400K context, 128K max output, avg@3 over three rollouts per task. |
| DeepSWE (v1.1) | 66.9 | mini-swe-agent harness, temperature 0.95, top_p 1.0, 6h timeout, 400K context. Pass rate type not stated. |
| CyberGym | 84.5 | Claude Code 2.1.207, max reasoning effort, no web tools, temperature 1.0, top_p 1.0, single-run Pass@1 over 1,507 tasks. |
| Toolathlon Verified | 73.0 | Official evaluation service, pass@1 averaged over 3 independent runs. |
How GLM-5.3 compares
GLM-5.3 is the newest Z.ai model in this catalog. It carries the coding tag, a 1M context, and lists at $1.40 per million input tokens and $4.40 output. Those are the same three facts as GLM-5.2, which means nothing recorded here separates the two except release order. The price difference is larger with GLM-5.3 Flash: the same 1M window and family, but $0.15 and $0.50 at list and tagged for vision rather than coding. Compare Flash on your own tasks to see whether its lower rates reduce the cost of accepted results. Against GLM-5 and GLM-5.1 the difference is window size, 1M against 200K.
These comparisons use catalog context limits, tags and reference rates, not independent quality benchmarks. Rates can change. Check current pricing above and test your workload; a higher price or a larger context window does not establish better results.
| Model | Context | Input / output per 1M | What differs |
|---|---|---|---|
| GLM-5.2 | 1M | — | Same base model and the same Z.ai list price. Z.ai says every GLM-5.3 gain comes from post-training. On its own model card Z.ai reports Terminal Bench 3.0 of 28.3 against 4.6 for GLM-5.2. |
| GLM-5.3 Flash | 1M | — | Z.ai calls it the first natively multimodal GLM-5 model, with image and video input. GLM-5.3 takes text only. On the Z.ai list on 2026-11-06 Flash input costs $0.15 per 1M tokens against $1.4 for GLM-5.3. |
| Qwen3.7 MAX | 1M | — | Both are agent-oriented flagships with a 1M window. GLM-5.3 always reasons; Qwen3.7-Max offers thinking and non-thinking modes. Compare the live rates and test your own tasks. |
Quick setup
curl https://a2agent.me/v1/chat/completions \
-H "content-type: application/json" \
-H "authorization: Bearer sk-your-a2agent-key" \
-d '{
"model": "glm-5.3",
"messages": [{"role": "user", "content": "Reply with OK"}]
}'Using it from a coding client
Reach for 5.3 when a coding agent has to read broadly before it writes. If the session is mostly retrieval and summarizing, GLM-5.3 Flash covers the same window for about a tenth of the money.
Questions
Is GLM-5.3 the newest GLM here?
Yes. It is the latest Z.ai release in this catalog, alongside GLM-5.3 Flash.
How is it different from GLM-5.2?
By the data shown here, it is not: same 1M window, same coding tag, same $1.40 and $4.40 list. GLM-5.3 is the later release.
What is the context window?
1M tokens.
When should I use GLM-5.3 Flash instead?
When the work is reading rather than writing. Flash keeps the 1M window and lists at $0.15 and $0.50, about a tenth of this rate.
Billed per token. Top up from $5, no top-up fee.
