This guide uses source checks from Sep 7, 2026. Provider and gateway prices can change; preserve the checked date with every forecast.
Why This Topic Matters Now
`glm 5 api` is a narrow query, but it is exactly the kind of query a Tier 1 developer uses when moving from curiosity to implementation. The Sep 6 keyword report recorded two impressions for `glm 5 api`, both from Tier 1 and led by the United States. Recent AIWave posts have already covered GLM migration, promotion expiry controls, and broader add-on billing. This article takes a narrower route-receipt angle: how to prove that a GLM-5 route works through an OpenAI-compatible gateway before customer traffic depends on it.
A route receipt is more useful than a screenshot of a pricing table. It records the model string, account group, source price row, checked date, request timestamp, response status, usage object, data boundary, and fallback policy. With that evidence, engineering can debug route changes, finance can review token classes, and security can confirm that the first test used redacted data. This is the right level of proof for a GLM-5 API evaluator who has not yet committed production traffic.
Source Facts Checked Today
AIWave /api/pricing checked on Sep 7, 2026 returned success=true, 63 records, pricing_version a42d372ccf0b5dd13ecf71203521f9d2, default group ratio 3, and VIP group ratio 1. The live pricing page checked the same day listed GLM-5 at $1.55 input, $0.4000008 cache-hit input, and $4.96 output per 1M text-token units. It listed GLM-5.1 at $2.10 input, $0.680001 cache-hit input, and $6.60 output.
The same AIWave public page checked on Sep 7, 2026 also lists adjacent GLM rows: GLM-4.5 at $0.6975 input, $0.1800003 cache-hit input, and $2.17 output; GLM-4.5-air at $0.6525 input, $0.1620001 cache-hit input, and $2.03 output; GLM-4.7 at $0.93 input, $0.2199999 cache-hit input, and $3.41 output; and GLM-5-turbo at $1.80 input, $0.4800006 cache-hit input, and $5.40 output.
The Sep 6 keyword report says the Z.AI pricing page checked during the Sep 7 run lists GLM model rows plus built-in tool, image, video, audio, and agent pricing sections with input, cached-input, cached-input-storage, and output columns. That provider context is useful, but a gateway receipt should label AIWave rows separately from direct-provider rows so a buyer does not confuse provider context with the current gateway account's effective price.
Planning Matrix
A source-dated planning matrix keeps the page useful for engineers and procurement reviewers. It turns a search query into an auditable route decision instead of a loose model preference.
| Receipt field | Why it matters | Example value |
|---|---|---|
| Model string | Catches alias drift | glm-5 |
| Source checked date | Keeps price claims current | 2026-09-07 |
| Account group | Explains multiplier | VIP or default |
| Input tokens | Separates request size | Recorded from usage object |
| Cache-hit tokens | Shows repeated-prefix behavior | Ledger field or exception note |
| Output tokens | Controls response budget | Recorded from usage object |
| Fallback rule | Prevents silent behavior change | GLM-only, or approved fallback pair |
Implementation Pattern
The implementation pattern keeps credentials as placeholders, pins the AIWave base URL, records the model, and leaves room for route-specific controls. Production applications should move credentials into environment or secret storage.
from openai import OpenAI
from datetime import datetime, timezone
client = OpenAI(api_key="YOUR_API_KEY_HERE", base_url="https://aiwave.live/v1")
def capture_glm_receipt(prompt: str) -> dict:
checked_at = datetime.now(timezone.utc).isoformat()
response = client.chat.completions.create(
model="glm-5",
messages=[{"role": "user", "content": prompt}],
max_tokens=250,
temperature=0.1,
)
return {
"model": "glm-5",
"source_checked_at": "2026-09-07",
"pricing_version": "a42d372ccf0b5dd13ecf71203521f9d2",
"request_checked_at": checked_at,
"usage": response.usage,
}
print(capture_glm_receipt("Summarize this redacted GLM route test in one sentence."))
Define the Receipt Before the Test
A GLM-5 route test should begin with a receipt schema. Decide which fields are required before the first request: model string, source URL, checked date, pricing_version, account group, prompt class, input tokens, cache-hit tokens when visible, output tokens, status, latency, and stop reason. If the team waits until after the test to decide what evidence matters, it usually keeps only the answer text and loses the operational proof.
Use Redacted Inputs
A first GLM-5 gateway test should use a redacted prompt that still reflects the real task. For example, use a synthetic support ticket, a public code snippet, or an anonymized extraction sample. The goal is to test request shape, route availability, output quality, usage fields, and billing evidence without sending sensitive production data. Security can approve a broader data class later if the first receipt is clean.
Pin Account Group and Price Row
The AIWave pricing page exposes base values and explains that the account group affects the effective price. A receipt should therefore include the account group that applied to the request. If the buyer tests with a funded VIP account, label it. If the buyer tests before VIP is effective, label the default group. That single field prevents finance from comparing a default-group receipt against a VIP base-rate forecast.
Watch Cache and Output Separately
GLM rows on AIWave separate input, cache-hit input, and output. Keep those fields separate in the receipt. A prompt with a stable system block may behave differently from a prompt that changes every call, and a short extraction answer behaves differently from a long reasoning answer. Output caps belong in route configuration, and cache assumptions belong in a measured prompt-template record.
Separate Direct Provider Context
Z.AI pricing is direct-provider context, while AIWave pricing is gateway context. Both can appear in a procurement packet, but they should not be merged into one unlabeled row. A provider row can explain upstream feature economics, tool costs, or cached-input structure. The AIWave row tells the buyer what the gateway page published for the tested route and account group. Keep row owner, source URL, and checked date in every table.
Internal Links for GLM Evaluators
A GLM evaluator should move through Models docs, Chat Completions, Pricing, Trust, the GLM promotion-expiry controls, and the add-on billing ledger. Those links give the reader pricing, implementation, trust, and adjacent billing context.
Procurement Review
Procurement should ask for one GLM receipt per approved route. Each receipt should include model string, source URL, checked date, pricing_version, account group, input tokens, cache-hit field status, output tokens, prompt data class, and fallback policy. Engineering should attach the redacted prompt class and the acceptance result. Finance should preserve AIWave and Z.AI rows separately so later price movement can be audited.
Final Checklist
A GLM-5 route is ready for a small trial when the model string is verified, the first request returns usage evidence, the price row is source-dated, the account group is explicit, cache and output are separated, the data boundary is approved, and the fallback rule is written down. Recheck AIWave pricing and Z.AI context before production traffic or monthly budget approval.