This guide uses source checks from Aug 30, 2026. Provider and gateway prices can change; preserve the checked date with every forecast.
Why This Topic Matters Now
The Aug 29 keyword report makes the brand problem concrete: `site:aiwave.live` had 94 impressions and no clicks, `aiwave.live` had 59 impressions and one click, and the United States continued to dominate Tier 1 visibility. The searcher is often not asking for a slogan. They are trying to decide which result proves that the model catalog, route names, pricing, and documentation are current enough for a trial.
The Aug 29 market research points in the same direction. It says one of the clearest supply gaps is context and version metadata: aliases, snapshots, output limits, endpoints, features, and change history. This article turns that gap into an SEO page and an operating checklist. It uses AIWave live metadata checked on Aug 30, 2026 and keeps public breadth language separate from machine-readable route counts.
Source Facts Checked Today
AIWave /api/pricing checked on Aug 30, 2026 returned success, 63 records in the data array, pricing_version a42d372ccf0b5dd13ecf71203521f9d2, supported_endpoint openai, default group ratio 3, VIP group ratio 1, and model records that expose model_name, quota_type, model_ratio, completion_ratio, cache_ratio where applicable, enable_groups, and supported endpoint types. Those fields are useful audit evidence, but the public marketing count remains the approved 25+ Chinese model routes.
The live feed includes examples across DeepSeek, Kimi, GLM, Qwen, ERNIE, MiniMax, Hunyuan, and other Chinese provider families. Current parsed rows include deepseek-v4-pro, deepseek-v4-flash, kimi-k3, glm-5.1, and qwen-72b-chat. Some records include a per-record pricing_version while others rely on the top-level version, so a catalog audit should store both the top-level version and the record-level fields that were visible at the time of evaluation.
Alibaba Model Studio documentation checked for this project documents region-specific endpoints and model availability behavior. That source is useful because it shows why a gateway catalog needs more than a model label. A Tier 1 buyer may need endpoint family, provider owner, region assumption, feature support, output limit, cache behavior, and billing basis before a route can move from experiment to production configuration.
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.
| Catalog field | Why it matters | Buyer evidence |
|---|---|---|
| Model name | Controls the request route | Pinned ID in configuration |
| Provider family | Affects policy and support path | Owner label and route note |
| Endpoint support | Prevents SDK mismatch | openai support flag |
| Pricing version | Anchors a forecast | Top-level and record version |
| Group support | Explains account eligibility | default, vip, svip flags |
| Cache ratio | Changes long-context cost | Visible cache field when present |
| Change date | Protects against alias drift | Checked date and release note |
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.
import requests
AIWAVE_API_KEY = "YOUR_API_KEY_HERE"
resp = requests.get(
"https://aiwave.live/api/pricing",
headers={"User-Agent": "AIWave catalog audit", "Authorization": f"Bearer {AIWAVE_API_KEY}"},
timeout=20,
)
payload = resp.json()
routes = []
for row in payload.get("data", []):
if "openai" in row.get("supported_endpoint_types", []):
routes.append({
"model": row.get("model_name"),
"groups": row.get("enable_groups", []),
"cache_ratio": row.get("cache_ratio"),
})
print(payload.get("pricing_version"), routes[:5])
Separate Public Count From Machine Data
Public copy should stay stable and approved: 25+ Chinese model routes. Machine-readable catalog data can be richer and more volatile. Mixing those two layers creates problems. If a live feed has 63 route records, that does not mean every public page should claim 63 distinct production-ready models. Some records may be variants, provider aliases, endpoint-specific routes, or operational rows. Use the public count for positioning and the API feed for engineering evidence.
Audit Route Names Before a Trial
A trial should begin by listing candidate model IDs from the models or pricing feed, then pinning a small set in configuration. Do not let prompt text choose the model. Do not let a UI dropdown become the source of truth for a production route. Store the route name, provider family, endpoint type, pricing_version, checked date, and acceptance result. That record is enough to explain why a team used one route in a funded test.
Watch Alias and Version Drift
Model aliases can change while the integration string stays stable. That can be good when a provider upgrades a route deliberately, but it is risky when quality, context size, output size, or pricing changes silently. A catalog audit should store visible model name, provider family, source date, pricing_version, and any documented snapshot or version label. When an alias moves, the team can run acceptance tests again instead of discovering drift from a support ticket.
Make Endpoint Support Visible
OpenAI-compatible does not mean every feature behaves identically. The catalog should show which routes support the OpenAI-style endpoint used by the application and which features were accepted in the local test. Streaming, tool calls, JSON response format, long output, and vision should be verified separately when a workflow depends on them. A field in the catalog is a starting point; the acceptance test is the launch gate.
Use Group Fields Carefully
The live AIWave feed exposes account-group information that matters for billing. Current project memory says VIP status applies to all tokens automatically and the default and VIP group ratios are distinct. Buyer-facing content should explain effective group as a billing field, not as something developers must manually set on tokens. The safest wording is that the account should verify its effective group in Console before a finance forecast.
Internal Links for Brand Searchers
Catalog readers should move through Models docs, Chat Completions, Pricing, Trust, Long Context API, and the streaming migration guide. That path answers the brand query with model evidence, endpoint evidence, rate evidence, and operational boundaries.
Procurement Review
Procurement should ask for the catalog snapshot, source URL, checked date, model ID, provider family, endpoint support, account group, and pricing_version. Engineering should attach one successful request for each candidate route and one failure-classification note for an unsupported feature. This is a narrow review, but it catches the main ways a model catalog can be too vague for production use.
Final Checklist
A catalog audit is ready when the public count stays approved, live route records are source-dated, model IDs are pinned, endpoint support is visible, pricing_version is captured, group fields are understood, cache fields are not hidden, and alias drift has a review path. That turns `site:aiwave.live` and model-doc searches into a credible first technical visit.