Old tiers retired: Worry-free and Pro are gone
Xunfei Astron Token Plan has retired the ¥19/mo Worry-free and ¥39/mo Pro tiers, leaving only the ¥199/mo High-efficiency and ¥999/mo Speed tiers on sale. Open the Astron MaaS subscription page (source) and you'll see only two cards under "choose your plan"—High-efficiency and Speed—while Worry-free and Pro have disappeared.
The essence of this change is cutting the cheap tiers and raising the entry bar. Worry-free used to offer "unlimited requests" but only lightweight models; Pro gave roughly 18,000 requests per month. Both are now replaced by High-efficiency and Speed. For existing users, the immediate impact is that the ¥19 and ¥39 options are gone at renewal.
Why the cheap tiers got cut
- Billing model changed: from pure request counts to request counts plus model deduction coefficients, which a cheap unlimited tier can't sustain.
- Model pool upgraded: hosting flagships like GLM-5.2, DeepSeek-V4-Pro and Kimi-K2.7-Code makes the low-price tier's marginal cost unworkable.
- Clearer positioning: Xunfei wants this subscription to be a "power developer's compute pack," not a student practice tier.
Two-tier pricing comparison
The pricing comparison is straightforward: High-efficiency at ¥199/mo is the high-volume workhorse, while Speed at ¥999/mo (first month ¥699) is the queue-free, all-model top tier. The core difference is quota and model coverage (source):
| Item | High-efficiency | Speed |
|---|---|---|
| Monthly | ¥199/mo | ¥999/mo (first ¥699) |
| Quarterly | ¥538/qtr (~¥179.3/mo) | ¥2,697/qtr (first) |
| Per 5 hours | 6,000 requests | Speed quota |
| Weekly | 45,000 requests | — |
| Monthly | 90,000 requests | ~30,000 requests |
| Highlight | 19-model pool | Queue-free, 2M TPM, all models |
The pricing comparison conclusion: High-efficiency gives more monthly requests (90,000 vs 30,000), but Speed buys priority response and full model coverage—for example Spark X2 Agent and Kimi-K2.7-Code are reliably available on Speed. Whether you want "run more" or "run faster" decides the tier.
Model deduction coefficients: flagships burn quota faster
Since July 23, 2026, new High-efficiency and Speed purchases apply model deduction coefficients, so higher-coefficient models consume more requests per prompt. This directly affects your effective quota (source):
| Model | Coefficient |
|---|---|
| GLM-5.2 / DeepSeek-V4-Pro / Kimi-K2.7-Code | x5 |
| Spark X2 / Spark X2 Agent / Auto | x2 |
| Spark-X2-Flash | x1 |
The point of this pricing comparison: the same 6,000 requests/5h lets you run 6,000 Spark-X2-Flash calls, but only 1,200 GLM-5.2 or DeepSeek-V4-Pro calls. Flagships aren't off-limits—they just drain your quota faster.
Is upgrading from an old tier worth it?
For former Pro (¥39/mo) users, upgrading to High-efficiency (¥199/mo) is worth it mainly if you're a heavy user. High-efficiency delivers roughly 5× Pro's monthly quota plus GLM-5.1, Spark X2, Kimi-K2.6 and DeepSeek V4, which covers most power users. But ¥39 to ¥199 is a big jump, and light users would be overpaying.
If you only code occasionally, compare sideways instead: for a cheaper fixed-fee multi-model pack, look at Baidu Qianfan; for dedicated MiniMax, Kimi or Zhipu GLM subscriptions, open each separately. The core of this pricing comparison is to measure your real monthly usage before choosing a subscription over pay-as-you-go.
API key and integration
Xunfei Astron Token Plan uses a dedicated coding key that connects to Claude Code, Cursor and TRAE through the MaaS coding endpoint—it is not interchangeable with the open API inference key. Setup (source):
- OpenAI Base URL:
https://maas-coding-api.cn-huabei-1.xf-yun.com/v2 - Anthropic path:
https://maas-coding-api.cn-huabei-1.xf-yun.com/anthropic - Default model
astron-code-latest; switch the base model on the MaaS page (effective in 1–3 min). - Supports Claude Code, Cursor, OpenCode, OpenClaw, Codex CLI, TRAE and more.
The dedicated key is for interactive coding-tool use only—using it for backends or automation scripts is a violation.
Things to watch
-
Worry-free and Pro are retired. There's no cheap replacement; after your old plan expires you can only buy High-efficiency or Speed.
-
Speed's first month at ¥699 is a limited promo. Renewals return to ¥999/mo, so budget long-term at ¥999.
-
Model coefficients only apply to purchases after July 23. Old plans keep the original deduction rules until expiry.
-
Validity is 31 days, not a calendar month. It starts at payment and unused quota does not roll over.