// Halo
Qwen vs DeepSeek — how the two top Chinese models compare
The two strongest open Chinese models, compared on reasoning, coding, languages and price — and how to run either without an account.
Qwen vs DeepSeek is the interesting comparison inside the Chinese open-model wave — both are open-weight, both are cheap, and both are genuinely competitive with frontier models on the tasks they’re good at.
At a glance
| Qwen | DeepSeek | |
|---|---|---|
| Reasoning / code | Strong; coder variants competitive | The line’s headline strength |
| Multilingual | Particularly broad coverage | Strong |
| Model sizes | Wide range, small to large | Focused line |
| Openness | Open weights | Open weights |
| On Halo | Pay per prompt, no account | Pay per prompt, no account |
How to choose
Reach for DeepSeek when the job is reasoning or code and you want the most capability per unit of cost. Reach for Qwen when you need breadth — a range of model sizes to pick from, or serious multilingual work beyond English and Chinese.
Both are open-weight, which is the quiet advantage over closed frontier models: nobody can withdraw them, and anyone with hardware can serve them. That’s what keeps the price competitive on a market like Halo, where operators bid for each request.
Access, without the phone number
The official apps for both tend to want a Chinese phone number or an account. On Halo neither applies — sign in with a social account and prompt. Nothing ties the conversation to a provider account.
Start with Qwen or DeepSeek, see the wider roundup of Chinese ChatGPT alternatives, or compare against a Western model in DeepSeek vs ChatGPT.
Frequently asked questions
Is Qwen better than DeepSeek?
They're close, and they split by task. DeepSeek's reputation is built on reasoning and code at very low cost; Qwen's line is broader, with particularly strong multilingual coverage and a wide range of sizes. Both are open-weight, so the useful test is your own prompts.
Which is better for coding?
DeepSeek is the more common pick for reasoning-heavy and coding work, though Qwen's coder-oriented variants are competitive. Since both cost a fraction of a frontier model on Halo, running the same task through each is cheap enough to be worth doing.
Do I need a Chinese phone number for either?
Not on Halo. The official apps often want a Chinese phone number or account; here you sign in with a social account and send a prompt to whichever model you want. No wallet needed, no provider signup.
Are they cheaper than Western models?
Usually by a wide margin, and that's most of their appeal — comparable reasoning at a small fraction of the cost. On Halo you pay per prompt in USDC on Base for either, so the saving shows up directly.