Global AI Token Comparison Hub
CN
Back to all plans

百度千帆

Aggregated models

Qianfan Coding Plan — GLM, Kimi, DeepSeek, and ERNIE Turbo in one coding subscription

Coding PlanCoding Plan
SubscriptionToken API
Official

Core models

GLM-5GLM-5.1Kimi-K2.5MiniMax-M2.5DeepSeek-V3.2DeepSeek-V4-ProDeepSeek-V4-FlashERNIE-4.5-Turbo-20260402
GLM-5

Zhipu GLM-5 flagship in the pool—for complex reasoning and code; peak hours deduct 3 quota calls per request, off-peak 2—factor multiplier into planning.

GLM-5.1

Newer Zhipu GLM-5.1 trial tier with limited capacity; peak 4× / off-peak 3× quota deduction—switch to ERNIE Turbo or DeepSeek-V4-Flash if queued.

Kimi-K2.5

Moonshot Kimi-K2.5 with 1:1 quota deduction—for conversational coding and long-file context; requires Coding Plan dedicated key, not ordinary Qianfan API keys.

MiniMax-M2.5

MiniMax M2.5 with 1:1 deduction—official docs mark it as sunsetting soon; new work should switch to Kimi-K2.5 or ERNIE-4.5-Turbo-20260402.

DeepSeek-V3.2

Value DeepSeek-V3.2 with 1:1 deduction for daily engineering—do not use plan keys for backend pay-as-you-go services.

See the official site for more models

Additional core model names still appear above, with full details on the latest official page.

Open official site

Plan details

Lite

Basic
Price
¥40/mo
1,200/5h · 9,000/wk · 18,000/mo calls
Official
Usage
At ¥40/mo for baseline development: ~1,200 calls per 5 hours, 9,000 weekly, 18,000 monthly—all three caps apply together, suited to medium use rather than nonstop extreme workloads.
Models
Switch among GLM-5, GLM-5.1, Kimi-K2.5, MiniMax-M2.5, DeepSeek-V3.2/V4, and ERNIE-4.5-Turbo to compare coding styles and debugging at low entry cost.
Highlights
5-hour window rolls with request time; weekly quota resets Monday 00:00 UTC+8; monthly quota resets on subscription renewal day 00:00 UTC+8—when exhausted, wait or upgrade to Pro.
Best with one terminal; suited to entry developers, script edits, routine debugging, and light agent work—a natural way to learn which model family you prefer before upgrading.
Best for
Entry developers, single-terminal daily users, and people wanting a low-barrier way to explore multiple models

Pro

ProfessionalRecommended
Price
¥200/mo
6,000/5h · 45,000/wk · 90,000/mo calls
Official
Usage
At ¥200/mo for complex, sustained work: ~6,000 calls per 5 hours, 45,000 weekly, 90,000 monthly—5× Lite quota for teams treating AI coding tools as primary productivity.
Models
Same 8 aggregated models with headroom for switching, long tasks, and complex debugging; Lite-to-Pro upgrades add prorated quota for the rest of the billing month.
Highlights
Recommend at most 2 terminals; suited to complex projects, high-frequency individual development, long-context debugging, and workflows that truly fold multiple models into the daily workbench.
Peak hours may queue or rate-limit—stagger usage, switch models, or use qianfan-code-latest auto-routing; parallel OpenClaw instances may trigger cluster rate limits.
Best for
Developers on complex projects, high-frequency AI coding users, and professionals using multi-model subscriptions as a main workbench

Quota

Product type
Coding Plan
Official product line
Coding Plan
5-hour limit
1,200–6,000 calls/5h
Weekly limit
9,000–45,000 calls/wk
Monthly limit
18,000–90,000 calls/mo

Notes

  • Purchases are non-refundable; Lite can upgrade to Pro with prorated difference, but Pro cannot downgrade to Lite.
  • Quota is for interactive use inside coding tools only—no automation scripts, app backends, or non-interactive batch calls; requires a dedicated Coding Plan API key and Base URL.
  • Lite recommends ≤1 terminal, Pro ≤2; multi-terminal or peak-hour use may hit rate limits. GLM-5.1 and DeepSeek-V4-Flash are trial tiers with limited capacity; MiniMax-M2.5 is marked as sunsetting soon—prefer Kimi or ERNIE Turbo for new setups.
  • First-buy promos (Lite ¥7.9, Pro ¥39.9) sold out per official notice on 2026-03-16; budget long-term at Lite ¥40/mo and Pro ¥200/mo.
  • Peak hours (~10:30–12:00, 14:00–18:00) may throttle some models—switch to 1:1 models like ERNIE Turbo or DeepSeek-V4-Flash, or use off-peak hours.

Supported coding tools

Claude CodeCursorClineOpenClawOpenCodeCodeXQwen Code Companion

Pricing and model data sourced from official vendor websites

FAQ

General·9
General
9 条