China-family models

Call GLM / Kimi / DeepSeek / Qwen with the same QCode.cc cr_ key: protocol path, live ids, curl check, and how they sit next to Claude 5

China-family models ¶

QCode.cc is not only Claude / GPT / Gemini. The same cr_ key also reaches Zhipu GLM, Moonshot Kimi, DeepSeek and Qwen. This page is only “how to hit these ids”, not a bake-off and not marketing.

The daily default is still the Claude 5 line (see Model Selection). Use the China-family tier for Chinese long-form work, cheaper bulk jobs, or a side-by-side check against Claude. Do not type a name that is missing from qcode.cc/models.

What is on sale ¶

Ids below were extracted from dev.qcode.cc/models on 2026-08-16 (positive control: the set contains claude-sonnet-5). Prices are the input / output rates per million tokens on that page at the time. Treat qcode.cc/models as live.

Family	Flagship id	Lighter sibling	Flagship snapshot (in / out)	Flagship context / out
Zhipu GLM	`glm-5.2`	`glm-5.1`	$1.40 / $4.40	200K / 131K
Moonshot Kimi	`kimi-k3`	`kimi-k2.6`	$3.00 / $15.00	1M / 131K
DeepSeek	`deepseek-v4-pro`	`deepseek-v4-flash`	$0.43 / $0.87	1M / 384K
Qwen	`qwen3.7-max`	`qwen3.7-plus`	$2.50 / $7.50	1M / 131K

Lighter-tier snapshot: kimi-k2.6 $0.95 / $3.96; deepseek-v4-flash $0.14 / $0.28; qwen3.7-plus $0.40 / $1.60. glm-5.1 matched glm-5.2 in/out that day, with a smaller window (81K).

No grok. Live list: qcode.cc/models or GET https://api.qcode.cc/v1/models with your key.

Which protocol ¶

The path selects the protocol; the key does not care which family you call. Rules: Endpoints & API Formats.

Tool	What to fill	Model field
OpenAI SDK / LangChain / Cline “OpenAI Compatible” / WorkBuddy	`https://api.qcode.cc/openai/v1` (in CN prefer `https://asia.qcode.cc/openai/v1`)	An id from the table, character-for-character
Claude Code (already pointed at QCode per Environment)	Leave `ANTHROPIC_BASE_URL` alone	`/model glm-5.2` in the session (same pattern as `/model gpt-5.6-terra`)
Raw HTTP	`POST /openai/v1/chat/completions`	JSON `"model"`

Do not send these ids to the Gemini /gemini prefix. Do not put https://api.qcode.cc/api into an OpenAI-compatible client (that prefix is Anthropic Messages).

Non-Claude models in Claude Code may lack native extras such as Extended Thinking — Model Selection already says the same for GPT. Treat them as a complement, not a new daily default, if you are unsure.

curl check ¶

Same reading as the endpoints page: 400 = path and key reached the gateway (empty body expected), 401 = bad key, 404 = wrong prefix.

KEY="cr_你的密钥"

# 只验路径
curl -s -o /dev/null -w '%{http_code}\n' -X POST https://api.qcode.cc/openai/v1/chat/completions \
  -H "Authorization: Bearer $KEY"

# 带一个在售国产 id 做端到端（把 glm-5.2 换成你要用的 id）
curl -sS https://api.qcode.cc/openai/v1/chat/completions \
  -H "Authorization: Bearer $KEY" \
  -H "content-type: application/json" \
  -d '{"model":"glm-5.2","messages":[{"role":"user","content":"ping"}]}'

In mainland China, repeat with host asia.qcode.cc. One key on all four domains.

Reference OpenAI Python client (base_url stops at /openai/v1, no trailing slash):

from openai import OpenAI

client = OpenAI(
    api_key="cr_你的密钥",
    base_url="https://api.qcode.cc/openai/v1",
)
resp = client.chat.completions.create(
    model="glm-5.2",
    messages=[{"role": "user", "content": "ping"}],
)

Next to Claude 5 ¶

Job	Start with	When to add a China-family id
Daily code edits	`claude-sonnet-5`	A comparison run, or `deepseek-v4-pro` / `glm-5.2` if budget is tight
Hard reasoning / large refactors	`claude-opus-5`	Optional; try `kimi-k3` / `qwen3.7-max` to save
Chinese long-form, notes, office	any flagship	`kimi-k3`, `glm-5.2`, `qwen3.7-max` are common
Bulk / chores	`claude-haiku-4-5`	`deepseek-v4-flash` / `qwen3.7-plus` were cheaper that day

4.x (claude-sonnet-4-6, claude-opus-4-8) is still on sale; it is just no longer the default. Rates and windows: qcode.cc/models. When this page was written, claude-sonnet-5 input was cheaper than claude-sonnet-4-6.

Same key on the desktop: WorkBuddy. Switching Claude Code / Codex: CC Switch Setup.

FAQ ¶

I filled the name and get 404 / an empty picker ¶

The id must look like glm-5.2, not “GLM-5.2” or a display label. OpenAI-compatible pickers are often empty — type it.

401 ¶

Key starts with cr_, no spaces. Check qcode.cc/dashboard.

Can I hit GLM on Anthropic `/api/v1/messages`?¶

The gateway picks the protocol from the path. /openai/v1/chat/completions is the path this site has verified for OpenAI-compatible clients. If Claude Code already points at QCode, switch with /model <id>; do not hand-roll Messages just to reach a China-family id.

WorkBuddy / Cline ¶

WorkBuddy: provider Custom, endpoint https://api.qcode.cc/openai/v1, model name = an id from the table. Steps: WorkBuddy Integration. Cline “OpenAI Compatible” uses the same Base URL and Model ID.

Next steps ¶

← Previous

Model Selection Guide

Permission Configuration

China-family models

Related Documents