Back to Models

GLM Search Pro

zhipuzhipu/search-pro

Flagship BigModel search β€” richer snippets, deeper coverage.

How TheRouter serves this differently from the vendor

As the vendor operates it

BigModel operates search_pro through its tools API as a web-search engine, not as a normal chat-completions model.

On TheRouter

TheRouter exposes it as zhipu/search-pro on /v1/chat/completions, translating the tools response into markdown result lines, url_citation annotations, and usage.web_search_requests billing metadata.

search-pro is the flagship web-search engine from Zhipu BigModel. It delivers richer content snippets, deeper crawl coverage, and better long-tail retrieval than search-std, while staying inside an OpenAI-style chat completion envelope so it drops into existing pipelines.

Pair it with a small chat model (glm-4.5-air, gpt-5-mini, claude-haiku) to build retrieval-augmented answers without managing your own crawler or embedding store.

Richer snippets
Per-result content fields are longer than search-std, surfacing the answer body directly inside the annotation.
Filtered queries
`search_recency_filter`, `search_domain_filter`, and `content_size: high` are accepted as standard fields.
Citation-ready output
Each hit yields a `url_citation` annotation with byte-accurate `start_index` / `end_index` into the markdown body.
Per-request billing
$0.0108 per call. Predictable cost independent of result size.
When to use
Production retrieval grounding for RAG, research agents, and answer engines that need solid coverage and tunable filters.
When not to use
If you only need link titles and don't care about snippet richness, search-std at ~3Γ— lower cost is enough.
Pricing: Flat $0.0108 per call. Filters do not add cost.
Read the web-search tutorial β†’
Context Length
8K
Max Output
--
Request Priceper request
$0.0054/ request

Modalities

text→text

Pricing Breakdown

TypeRate
Request$0.0054 / request

Supported Parameters

search_recency_filtersearch_domain_filtercontent_size

Specifications

Endpoint shapeOpenAI-compatible /v1/chat/completions alias backed by BigModel POST /api/paas/v4/toolsdocs.bigmodel.cn β†—verified
Search enginesearch_pro β€” BigModel self-developed premium multi-engine searchdocs.bigmodel.cn β†—verified
Supported filterssearch_recency_filter, search_domain_filter, content_sizetherouter.ai/guides/zhipu-web-search β†—verified
Response citationsmessage.annotations entries of type url_citation with URL, title, content, start_index, and end_indextherouter.ai/guides/zhipu-web-search β†—verified
Billing unitPer request via usage.web_search_requests; token counts are zero for search aliasestherouter.ai/guides/zhipu-web-search β†—verified
Streaming behaviorNo partial upstream streaming; stream=true returns one complete chunk followed by [DONE]therouter.ai/guides/zhipu-web-search β†—verified
Published vendor USD list priceNot published as a separate search-pro SKU on z.ai; TheRouter uses a platform per-request pricez.ai/pricing β†—verified

Benchmarks

BenchmarkDistributionScoreSource
Search retrieval quality
The public docs describe search-pro as richer and higher-recall than search-std, but no reproducible benchmark score is published.
β€”Not publicly disclosedβ€”
End-to-end retrieval latency
No public latency benchmark is available for TheRouter's search-pro alias; teams should measure regional retrieval latency before placing it on critical paths.
β€”Not publicly disclosedβ€”
Snippet/citation execution verification
The code snippets in this page have not been executed in the paid verification ledger. This is intentionally left unverified until an operator approves SNIPPET_VERIFY_APPROVED=1.
β€”Not publicly disclosedβ€”

API Usage Examples

Use the global api.therouter.ai endpoint shown below for new integrations; the legacy China accelerated endpoint is retired.

cURL
curl https://api.therouter.ai/v1/chat/completions   -H "Content-Type: application/json"   -H "Authorization: Bearer $THE_ROUTER_API_KEY"   -d '{
    "model": "zhipu/search-pro",
    "messages": [
      {"role": "user", "content": "Summarize the key points from this input."}
    ]
  }'

Search via chat completions

Call search-pro through the OpenAI-compatible chat-completions endpoint. Inspect message.annotations for url_citation entries and usage.web_search_requests for per-request billing.

cURL
curl https://api.therouter.ai/v1/chat/completions \
  -H "Authorization: Bearer $THEROUTER_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"model":"zhipu/search-pro","messages":[{"role":"user","content":"Summarize the latest Anthropic model news with sources"}],"search_recency_filter":"oneWeek","content_size":"high"}'

More from zhipu

Similar models

Cross-provider sibling models

News & changes

2026-05-20

TheRouter adds Zhipu Web Search routes

The launch added zhipu/search-pro alongside search-std, search-pro-sogou, and search-pro-quark as first-class /v1/chat/completions aliases with citation annotations and per-request billing.

2026-07-29

Zhipu search pricing reviewed against public rate cards

The rate-card sweep found that z.ai publishes only one USD Web Search SKU and no separate USD search-pro price; TheRouter therefore labels search-pro pricing as a platform per-request price rather than an official vendor USD list price.

re-authored by TheRouterz.ai/pricing β†—

Frequently asked

Is zhipu/search-pro a chat model?

No. It is a web-search alias exposed through the chat-completions envelope for SDK compatibility. Use it to retrieve cited web results, then pass those results to a separate model for synthesis if you need a polished answer.

How is search-pro different from search-std?

search-pro is positioned as the richer premium BigModel engine: deeper coverage, longer snippets, stronger recall, and support for the same recency/domain/content-size controls. Use search-std when basic low-cost search is enough.

re-authored by TheRouterdocs.bigmodel.cn β†—
Does stream=true give incremental search results?

No. BigModel's tools endpoint does not support partial streaming. TheRouter accepts stream=true for compatibility, but emits one complete result chunk and then [DONE].

What does TheRouter bill for a search-pro call?

The model is priced per request. Successful search calls include usage.web_search_requests: 1, and prompt/completion token counters are zero for this alias.

Fact ledger β€” every claim on this page traces here
sourceURLretrieved
Endpoint shapedocs.bigmodel.cn β†—2026-08-02verified
Search enginedocs.bigmodel.cn β†—2026-08-02verified
Supported filterstherouter.ai/guides/zhipu-web-search β†—2026-08-02verified
Response citationstherouter.ai/guides/zhipu-web-search β†—2026-08-02verified
Billing unittherouter.ai/guides/zhipu-web-search β†—2026-08-02verified
Streaming behaviortherouter.ai/guides/zhipu-web-search β†—2026-08-02verified
Published vendor USD list pricez.ai/pricing β†—2026-08-02verified
Search retrieval qualitydocs.bigmodel.cn β†—2026-08-02unknown
End-to-end retrieval latencytherouter.ai/guides/zhipu-web-search β†—2026-08-02unknown
Snippet/citation execution verificationtherouter.ai/guides/zhipu-web-search β†—2026-08-02unknown
TheRouter adds Zhipu Web Search routestherouter.ai/blog/zhipu-search-and-glm-4-5-air-launch β†—2026-08-02verified
Zhipu search pricing reviewed against public rate cardsz.ai/pricing β†—2026-08-02verified
Is zhipu/search-pro a chat model?therouter.ai/guides/zhipu-web-search β†—2026-08-02to verify
How is search-pro different from search-std?docs.bigmodel.cn β†—2026-08-02to verify
Does stream=true give incremental search results?therouter.ai/guides/zhipu-web-search β†—2026-08-02to verify
What does TheRouter bill for a search-pro call?therouter.ai/guides/zhipu-web-search β†—2026-08-02to verify
Help & contact