Qwen3 14B API
Model source: Hugging FaceThinks before answering, and you decide how much: a balance of quality and spend.
Qwen3 14B thinks before it answers, and you decide how much: a dial between quality and spend for tasks that need a few reasoning steps but not a frontier model. On GPU Flow it costs €0.20 per million input tokens and €1.00 per million output.
Reads in one go
≈ 70 pages
32,768 tokens
Brain size
Small and fast
14B, FP8
Where it runs
Spain (EU)
MIG B200
Understands
Text, not images
Chat, Reasoning, Tools
What it is good at
- Reasoning with adjustable effort. You set how much it thinks before answering, so you only pay for the reasoning the task needs.
- Tool calling for agents. Calls your functions in the standard OpenAI format and works with the result.
- Plain-language chat and summaries. Clear answers and summaries for everyday tasks, within a context of about 70 pages.
For most workloads Qwen3.8 27B is both stronger and cheaper on output; pick Qwen3 14B when you specifically want controllable reasoning effort.
Pricing
€ per million tokens, VAT included
For your technical team
Compatible with the OpenAI SDK: change the base URL and the key, and that is it.
curl https://api.gpuflow.ai/v1/chat/completions \
-H "Authorization: Bearer $GPUFLOW_API_KEY" \
-d '{ "model": "qwen-3-14b",
"messages": [{ "role": "user", "content": "Hola" }] }'Context and limits
- Context
- 32,768 tokens per request, about 70 pages.
- Requests per minute
- 1,000 per key, with bursts up to double; raisable to 5,000 from the dashboard. 5,000 in total per account.
- Tokens per minute
- 2,000,000 input (cached tokens do not count) and 400,000 output, per key.
- Prompt per request
- Up to 256,000 tokens, raisable on request.
- If you go over
- A 429 error with a Retry-After header telling you how many seconds to wait.
Qwen3 14B compared with similar models
| Qwen3 14B | Qwen3.8 27B | Qwen3.5 9B | |
|---|---|---|---|
| Best for | Reasoning with adjustable effort, tools | Coding, agents, structured output | Simple tasks at high volume |
| Context | ≈ 70 pages | ≈ 210 pages | ≈ 140 pages |
| Input / output, € per million | 0.20 / 1.00 | 0.10 / 0.30 | 0.10 / 0.15 |
| 1,000 chat answers | ~0.40 € | ~0.14 € | ~0.10 € |
Start with Qwen3.8 27B; choose Qwen3 14B when you want to control the reasoning effort; drop to Qwen3.5 9B if the task is simple and the volume is high.
Frequently asked questions
Is Qwen3 14B free?
The model weights are open source; running them costs GPU time. On GPU Flow you pay €0.20 per million input tokens, and every new account gets €5 of credit when it adds a card.
What does the Qwen3 14B API cost?
€0.20 per million input tokens, €1.00 per million output and €0.02 per million cached input. VAT included, no monthly fee.
Does Qwen3 14B support function calling?
Yes, with the standard OpenAI tools parameter.
Can I use Qwen3 14B with Claude Code or Cursor?
Yes. Claude Code connects through the Anthropic-compatible endpoint (/v1/messages); Cursor and other tools through the OpenAI-compatible one (/v1/chat/completions).
Where does it run?
On our NVIDIA B200 GPUs, in a data centre in Spain.
How do I set the reasoning effort?
With reasoning.effort ("low", "medium", "high", or "none" to switch it off) or with reasoning.max_tokens in the OpenAI SDK; with thinking.budget_tokens in the Anthropic one. We translate it to the engine for you.
When is Qwen3 14B better than Qwen3.8 27B?
When you specifically want to control how much the model reasons on each request. For most other workloads, Qwen3.8 27B is stronger and cheaper on output.