Hermes Agent + Qwen: Use Alibaba Qwen Models with Hermes

What Is Qwen?

Qwen (Tongyi Qianwen) is Alibaba's series of large language models, developed by the DAMO Academy. The Qwen3 generation brought major improvements in reasoning, coding, and instruction-following. Qwen models are particularly strong at Chinese language tasks and Asian multilingual use cases while remaining competitive in English and other major languages.

Hermes Agent supports Qwen through two paths: an OAuth flow via portal.qwen.ai (similar to the OpenAI Codex OAuth flow) or a plain API key via Alibaba's DashScope platform using the DASHSCOPE_API_KEY env var.

Available Qwen Models

Qwen 3.6 Plus is the recommended model for most Hermes + Qwen deployments. It handles a wide range of agent tasks well, with particularly strong performance on multilingual conversations and structured output generation. If you have Chinese language users or need solid East Asian language support, Qwen is often a better fit than Claude or GPT at similar cost.

Option 1: Hermes Agent on OpenClaw Launch (Easiest)

Qwen 3.6 Plus is available in the OpenClaw Launch model picker with no API key setup needed.

1. Go to openclawlaunch.com/hermes-hosting and start a Hermes deploy.
2. Select Qwen 3.6 Plus from the model dropdown.
3. Connect your channel and click Deploy. Your Qwen-powered Hermes Agent is live in roughly 10 seconds.

Option 2: DashScope API Key (Self-Hosted)

Alibaba's DashScope platform provides REST API access to Qwen models. Hermes reads the DASHSCOPE_API_KEY env var for this path:

export DASHSCOPE_API_KEY=sk-...

hermes inference set qwen
hermes model set qwen3.6-plus

# config.yaml equivalent:
# inference:
#   provider: qwen
# model:
#   default: qwen3.6-plus

Create a DashScope API key at dashscope.aliyun.com. DashScope billing is usage-based and priced in CNY; international credit cards are accepted.

Option 3: Qwen OAuth Flow (Self-Hosted)

Hermes also supports authenticating through the Qwen portal at portal.qwen.ai via an OAuth device flow, similar to how the OpenAI Codex flow works. Run the Hermes CLI authentication command and follow the prompts to authenticate through the Qwen website. This avoids managing a DashScope API key if you prefer browser-based auth.

Consult the upstream Hermes README for the exact CLI command in your installed version.

Option 4: Qwen via OpenRouter (Self-Hosted)

Qwen models are also available on OpenRouter if you prefer managing one key:

export OPENROUTER_API_KEY=sk-or-...

hermes inference set openrouter
hermes model set qwen/qwen3.6-plus

When to Choose Qwen over Other Models

Choose Qwen when your users primarily communicate in Chinese, Japanese, Korean, or other East Asian languages. The model's training distribution makes it produce more natural, idiomatic output in these languages than Western-primary models like Claude or GPT.

For English-primary use cases, Qwen 3.6 Plus is competitive with Claude Haiku and DeepSeek V4 Flash — strong general capability at moderate cost. If coding is the primary use case, compare it with GLM-5.1 and DeepSeek V4 Pro, which have higher published coding benchmark scores.

Switching to Qwen at Runtime

/model qwen/qwen3.6-plus

What's Next?

• Hermes Agent + GLM — Another strong Asian-origin model with SWE-Bench-leading coding scores
• Hermes Agent + DeepSeek — DeepSeek V4 Pro for cost-efficient frontier coding
• Hermes Agent + OpenRouter — Access Qwen and 200+ models with one OpenRouter key
• Hermes Agent + Telegram — Deploy your Qwen-powered Hermes bot on Telegram