Alibaba
Qwen3 4B
Qwen instruction model for multilingual chat, reasoning, and tool use
Model facts
- Context window
- 128K tokens
- Maximum output
- 33K tokens
- Cheapest paid input
- $0.03 per 1M tokens
- Cheapest paid output
- $0.03 per 1M tokens
- Open weights
- yes
- Providers
- 2
Capabilities and modalities
reasoning, tool calling, temperature control, open weights, text.
Local hardware estimate
At 4K context: Q4_K_M 3.4 GiB, Q8_0 5.3 GiB, F16 8.8 GiB working memory.
- Parameters
- 4.0 billion
- Architecture
- qwen3
- Model license
- apache-2.0
Planning estimate adapted from the MIT-licensed whichllm estimator using metadata from Hugging Face; not a vendor minimum requirement.
Artificial Analysis benchmarks
Compare this model on the full LLM benchmark leaderboard.
| Configuration | AAI | Coding | Math | Output tokens/s |
|---|---|---|---|---|
| Reasoning | 8.2 | not measured | 22.3 | not measured |
| Non-reasoning | 6.5 | not measured | not measured | not measured |
Observed price history
- 2026-08-24: $0.03 input / $0.03 output per 1M tokens
API providers
- NovitaAI — model id qwen/qwen3-4b-fp8; $0.03 input / $0.03 output per 1M tokens; provider documentation
- Pioneer — model id Qwen/Qwen3-4B-Instruct-2507; $0.2 input / $0.2 output per 1M tokens; provider documentation