Qwen Code Review, Pricing & Alternatives (2026)

Qwen Code is an open-source terminal coding agent maintained by Alibaba’s Qwen team, with its repository created on June 26, 2025, and licensed under Apache License 2.0. It began as a terminal interface for agentic programming and has expanded to editor plugins, desktop and web interfaces, SDKs, and messaging integrations. Developers run it with the `qwen` command to inspect repositories, edit files, execute commands, and use configured tools. Its provider-neutral design supports hosted services, compatible APIs, and local infrastructure rather than tying the client to one inference vendor. Qwen Code remains maintained after Qwen OAuth’s hosted free route ended on April 15, 2026. That change affected hosted access, while the software itself remains available under its existing license for modification, redistribution, and self-hosting.

The qwen command starts a repository session after the developer selects an authentication route. The agent can inspect files, propose a plan, edit code, and run approved commands or configured tools, while the developer reviews changes during the process. Project and user configuration resides in.qwen directories, with settings.json defining providers and environment variables supplying credentials. The /model picker allows runtime switching without restarting the terminal client. Qwen Code also documents auto-memory, skills, subagents, agent teams, MCP, custom commands, summaries, and local or remote endpoints. OpenAI-compatible support lets a team point the same interface at hosted APIs, proxies, or local servers. That arrangement makes the client useful for repository work across mixed infrastructure, since the terminal workflow stays consistent while authentication, provider selection, and model placement change.

Qwen Code supports multiple model routes rather than binding developers to one service. The terminal client works with QwenCloud, Alibaba Model Studio, OpenAI-compatible endpoints, Anthropic, Google Gemini, Vertex AI, and custom local servers. Its Qwen3-Coder lineup includes 256K-token native context, with Qwen3-Coder-Next, Qwen3-Coder-30B-A3B-Instruct, and Qwen3-Coder-480B-A35B-Instruct documented for coding. Alibaba describes the 480B model as a mixture-of-experts system with 480 billion total parameters and 35 billion active parameters per token. The project, created in 2025, added provider and billing changes during 2026, including QwenCloud Token Plans, Model Studio Coding Plan support, and broader authentication configuration. Developers can serve models locally through Ollama or vLLM, while the /model picker supports runtime switching among configured providers.

The client itself costs $0 under Apache License 2.0, but Qwen Code includes no universal hosted model allowance. Alibaba Model Studio’s current Coding Plan is Pro at $50/month, offering up to 6,000 requests per five hours, 45,000 per week, and 90,000 per month. These limits count model invocations rather than unlimited tokens, so a straightforward task can consume several requests. QwenCloud sells separate Personal Token Plans: Lite costs $8/month for 11,500 monthly Credits, Essential costs $16 for 25,500, Standard costs $25 for 45,000, and Pro costs $80 for 180,000. Limited-time prices are $6, $10, $18, and $68. Team seats cost $30, $100, or $200 monthly, with 25,000, 100,000, or 250,000 Credits respectively. BYOK users pay provider rates, while local users cover hardware and hosting.

Billing is Qwen Code’s main complication because quotas, retention, training terms, rate limits, and credentials vary by route. The Apache License 2.0 client costs $0, but hosted model access is separate: Alibaba Model Studio’s current Pro Coding Plan costs $50 per month and allows up to 6,000 requests per five hours, 45,000 per week, and 90,000 per month, with quota counting model invocations rather than tokens. QwenCloud separately lists Personal plans from $8 to $80 monthly and Team plans from $30 to $200 per seat. Claude Code provides a more tightly managed Anthropic route, Gemini CLI centers Google services, and Aider offers another provider-neutral terminal client. Qwen Code suits developers who prioritize Apache licensing, local Ollama or vLLM endpoints, broad provider choice, and Qwen3-Coder access over one predictable hosted bill.