Rating certificate · AW/26-09/52826 Sept 2026
Not enough evidence
The Register / Coding agents / Qwen Code
Register entry · Coding agents · updated 26 Sept 2026
Qwen Code
Qwen Code started as a fork of Gemini CLI tuned for Qwen models and now ships a CLI, desktop app, experimental web UI, VS Code/Zed/JetBrains integrations, chat-app channels and SDKs. It supports OpenAI, Anthropic, Gemini and Qwen API protocols plus local models via Ollama or vLLM. The hosted free OAuth tier ended on 2026-04-15, so model access now comes from Alibaba Cloud plans, third-party API keys or local inference.
What reviewers and users praise
- Works well with local Qwen models on consumer hardware (several HN reports, e.g. Qwen3-Coder-Next on M1 64GB)
- Model- and provider-agnostic, open source (README; HN users switching providers)
- Fast and solid on small real-world tasks (SwellJoe agent comparison, grade B-)
What they criticise
- Free hosted tier ended April 2026 (issue #3203)
- Authentication/401 errors and failed tool calls reported with some setups (HN)
- GitSpawn .git/config command-execution flaw was unpatched for about two months after report (The Hacker News, 2026-09-02); follow-up issues still open
Experts vs the crowdneeds both scores
Experts—
Crowd—
| Price | Free |
|---|---|
| Use it via | CLI, Desktop app, Web app, IDE extension, Chat apps, API, Browser extension |
| Runs on | macos, linux, windows |
| Licence | Apache-2.0 |
| Updated | Evidence collected 26 Sept 2026 |
How the rating was worked out
Weights for Coding agents: adoption 35%, experts 40%, crowd 25%. Fewer than two pillars have enough evidence. The method.
| Adoption & momentumweight 35% | 69 | GitHub stars 28k → 74 · npm weekly downloads 69k → 50 · VS Code / Open VSX installs 367k → 42 · 43 releases in 90 days → 100reach × 0.7 + momentum × 0.3 |
|---|---|---|
| Expertsindependent reviews · weight 40% | — | Only 1 independent review (3 needed) |
| Crowdcommunity sample · weight 25% | — | Community sample from a single platform |
| Rating | — | Not rated: fewer than two pillars have enough evidence. |
Evidence checks. 1 item removed: excerpt not found at the source (1). How evidence is checked.
Trust & safety
Seven dimensions graded A–D from documented facts. Overall: B.
C
Permission model
Default mode is 'Auto': an LLM classifier auto-approves actions it judges safe. Plan, Ask Permissions, Auto-Edit and YOLO modes are documented; sandboxing (bwrap execution sandbox, Seatbelt, Docker/Podman) is opt-in. Rubric: auto-executes by default, controls opt-in. source
B
Data access scope
Operates on the launched workspace with trusted-folders and .qwenignore settings documented; broad access unless the opt-in sandbox is configured. Browser extension can read tabs, cookies and history when used. source
A
Data storage
Runs locally; sessions and config stored on the machine; model calls go to whichever provider the user configures (or a local model). source
n/e
Data retention & training
The tool defers to the chosen provider's terms (Alibaba Cloud, OpenAI, Anthropic, Vertex, etc.); Qwen Code's own doc does not state a training or retention policy. Optional usage statistics described separately. source
C
Incident history (24 months)
GitSpawn: malicious .git/config (core.fsmonitor etc.) could execute code outside the sandbox without approval; Alibaba accepted the report 2026-07-07, still unpatched in 0.22.3 on 2026-09-02. Fix PRs merged 2026-08-08 (#8645) and 2026-09-12 (#11669), but issue #10193 'incomplete #8575 fix allows Git diff drivers to execute commands without approval' was still open on 2026-09-26. Graded C (serious, largely resolved); could be D if the open follow-up is judged unresolved. source
n/e
Compliance
No attestation published for Qwen Code; Alibaba Cloud certifications for Model Studio not checked.
A
Transparency
Apache-2.0 source with public GitHub releases and notes. source
The experts
1 named independent reviews found; dated reviews from the last 12 months are scored.
- positive
Quick, made one small mistake caught in code review. Just a solid model and a solid agent that didn’t bring attention to itself in any way
Joe Cooper (SwellJoe blog) · 3 Jul 2026 · swelljoe.com
The crowd
9 coded posts from hackernews.
5 positive0 mixed4 negative
- positive
I made this 3D game in a day on the same setup with Qwen Code as agent
news.ycombinator.com · 17 Sept 2026 · news.ycombinator.com - negative
Having audited one code harness (Qwen Code) in the past on telemetry. Let's say I'm sus of the Zuck offering free candy.
news.ycombinator.com · 6 Aug 2026 · news.ycombinator.com - negative
Also tried Qwen Code. Tried it with a coding plan and it 401'd. Tried it with an api key I loaded up with pay-as-you-go credits and it 401'd.
news.ycombinator.com · 3 Aug 2026 · news.ycombinator.com - negative
it often fails during `write_file` tool calls, even with Qwen Code.
news.ycombinator.com · 12 Jul 2026 · news.ycombinator.com - positive
one of my favorite things about local Qwen and Qwen code, it seems a heck of a lot faster that Claude and feels better to work with.
news.ycombinator.com · 17 May 2026 · news.ycombinator.com - negative
qwen code (non local) was the only time I managed to lose some work as it got confused when I ran out of tokens.
news.ycombinator.com · 12 May 2026 · news.ycombinator.com - positive
I've been using qwen-code (the software ...) which is a fork of gemini-cli and the tool use with Qwen models at least has been great.
news.ycombinator.com · 20 Apr 2026 · news.ycombinator.com - positive
In my experience Qwen3.5/Qwen3-Coder-Next perform best in their own harness, Qwen-Code.
news.ycombinator.com · 4 Mar 2026 · news.ycombinator.com
Pricing
As published, checked 26 Sept 2026.
| Plan | Price | Notes |
|---|---|---|
| Qwen Code software | Free | Apache-2.0; bring your own model access. |
| Qwen OAuth free tier | $0 (see notes) | Discontinued 2026-04-15; no longer selectable in /auth. |
| Alibaba Cloud Coding Plan (Pro) | Custom | Docs call it the 'official recommended setup'; weekly quota, multiple model families. Launched at ~$50/mo Pro (~$10/mo Lite) per Alibaba Cloud's X post; Lite closed to new subscribers 2026-03-20. Current USD price not verified on the vendor page. |
| Alibaba Cloud Token Plan | Custom | Usage-credit plan for teams; prices seen only in CNY on third-party pages. |
Recent changes
Ranked alongside
Other ranked entries in Coding agents.
OpenCodeAnomaly75
Claude CodeAnthropic73
Zed (Agent Panel)Zed Industries73
GitHub CopilotGitHub (Microsoft)72
Codex (CLI and app)OpenAI71
Sources
Every figure above traces to one of these, observed up to 26 Sept 2026.