Price terms as published
Each line is the vendor’s own wording, split from the price capture checked on 2026-10-16. Nothing here is rounded, converted or re-worded — the full sentences stay in the review below. Lines that carry only a price continue the plan named on the line above, in the vendor’s own order.
| # | Plan / term | Checked |
|---|---|---|
| 1 | Free (open source) | 2026-10-16 |
| 2 | token costs vary by provider | 2026-10-16 |
One-liner: Qwen Code is an open-source AI coding agent that runs directly in your terminal. Developed by Alibaba’s QwenLM team, it supports multiple model providers and is fully free — you pay only for the tokens you consume.
Verified facts (checked 2026-10-16)
| Attribute | Value | Source |
|---|---|---|
| GitHub stars | 28,509 | QwenLM/qwen-code via GitHub API |
| License | Apache-2.0 | LICENSE file, verified same day |
| Created | 2025-06-26 | GitHub API (over 1 year old) |
| Last push | 2026-10-16 | GitHub API (very active) |
| Supported providers | Qwen (native), OpenAI, Anthropic, Gemini, local Ollama, vLLM, and more | README auth docs |
| Runtime | Terminal / CLI | repo description |
| Price (agent shell) | $0 — Apache 2.0, fully open source | GitHub repo |
The README’s auth section is clear: /auth # Configure your provider and API key. You pick your upstream, plug in a key, and the token costs hit that provider’s billing.
What “free” actually means
The Qwen Code agent itself — the CLI tool, the planning logic, the tool-use framework — is completely free under Apache 2.0. You can self-host, modify, fork, and use it commercially with no restrictions beyond standard Apache attribution.
The cost is entirely on the model side:
- Qwen models (via DashScope / Bailian): Alibaba’s own models have had shifting free tiers historically. Free OAuth-style access has tightened before, and there’s no guarantee current free limits will stay.
- Third-party models (OpenAI, Anthropic, Gemini): Standard pay-as-you-go token pricing through each provider.
- Local models (Ollama, vLLM): Zero token cost if you run on your own GPU.
What the marketing claims don’t say
- “Free coding agent” ≠ free to use. Many roundups list Qwen Code as a “free Cursor alternative” without mentioning that every call costs tokens on whichever provider you connect. The free-part is the shell, not the intelligence.
- Qwen’s free tier history matters. Alibaba’s AI offerings (DashScope, Bailian) have a track record of promotional free tiers that get tighter, capped, or removed as adoption grows. If you’re building on the free Qwen path, assume the pricing can change.
- Apache 2.0 is permissive, but model provider terms aren’t. The agent code is yours to do anything with, but the models powering it are governed by each provider’s API terms — commercial use, rate limits, and data policies all come from whoever you plug in.
Who should use it
Teams that want a self-hosted, vendor-agnostic coding agent they can tailor to their stack; developers who prefer terminal-based workflows and want to choose their model provider; anyone running local models on their own hardware who wants a full agent framework on top.
Repository stats and feature claims verified against primary sources on 2026-10-16. This entry follows our verification policy: every claim was checked against primary sources, not copied from other reviews.