---

Features

• 6 text models — Qwen3.8 Max Preview, Qwen3.7 Max/Plus, Qwen3.6 Flash, GLM 5.2, DeepSeek V4 Pro
• Thinking support — enable_thinking for every model, with native reasoning_effort levels for GLM and DeepSeek
• Vision models — Qwen3.8 Max Preview, Qwen3.7 Plus, Qwen3.6 Flash accept image input
• 1M context — all models support up to 1M token context windows
• Zero config — just set your API key and go

Supported Models

Model	Context	Reasoning	Vision	Thinking Control
qwen3.8-max-preview	1M	✅	✅	on/off (enable_thinking)
qwen3.7-max	1M	✅	—	on/off (enable_thinking)
qwen3.7-plus	1M	✅	✅	on/off (enable_thinking)
qwen3.6-flash	1M	✅	✅	on/off (enable_thinking)
glm-5.2	1M	✅	—	7 levels (reasoning_effort)
deepseek-v4-pro	1M	✅	—	off, low, medium, high, xhigh, max

GLM 5.2 Thinking Levels

GLM 5.2 supports named reasoning effort levels mapped to pi's thinking selector:

pi Level	GLM reasoning_effort
off	none
minimal	minimal
low	low
medium	medium
high	high
xhigh	xhigh
max	max

Installation

Option 1: pi install (Recommended)

pi install npm:pi-qwencloud-token-plan-provider
# or from GitHub
pi install https://github.com/uttkarsh-26/pi-qwencloud-token-plan-provider

Option 2: Manual

git clone https://github.com/uttkarsh-26/pi-qwencloud-token-plan-provider.git
pi -e /path/to/pi-qwencloud-token-plan-provider

Authentication

Token Plan API keys start with sk-sp- (different from regular Qwen Cloud sk- keys).

Option 1: auth.json (recommended)

Add to ~/.pi/agent/auth.json:

{
  "qwen-cloud": {
    "type": "api_key",
    "key": "sk-sp-your-token-plan-key"
  }
}

Option 2: Environment variable

export QWEN_CLOUD_API_KEY=sk-sp-your-token-plan-key

Get your API key from the Token Plan console.

Plans

Plan	Price	5h Limit	7d Limit	Concurrency
Lite	$8/mo	700 Credits	2,500 Credits	1-2 agents
Standard	$25/mo	3,000 Credits	10,000 Credits	3-4 agents
Pro	$80/mo	12,000 Credits	40,000 Credits	6-8 agents

Usage

After installation, select a model with /model or:

pi --provider qwen-cloud --model qwen3.8-max-preview

Notes

• Costs show as $0 — Token Plan uses Credits-based billing, not per-token pricing
• The developer role is not supported; the extension sends system instead
• Token Plan is for interactive use only — no automated scripts or batch processing
• Region: Singapore (global inference)

Links

• Token Plan docs
• Thinking mode docs
• pi-coding-agent

License

MIT