Qwen3 14B

alibaba

Qwen3 hybrid reasoning model enables seamless switching between thinking and non-thinking modes during conversations. It achieves SOTA reasoning performance at its scale and significantly outperforms Qwen2.5-14B in general capabilities.

Try Now

Capabilities

Tool Use

Extended Thinking

Example Use Cases

Mid-size reasoning model needed

Moderate complexity with good efficiency

Self-hosted reasoning task

Technical Specifications

Context Window

131,072 tokens

Max Output

8,192 tokens

Cache Miss Cost

$0.35 per 1M tokens

Non-Reasoning Cost

$1.40 per 1M tokens

Reasoning Cost

$4.20 per 1M tokens

Web Search Cost

$15 per 1K calls

Code Execution Cost

$0.19 per 1K calls

⚠️ Legacy

Made legacy on

Reason

Untested

Recommended Replacement

Qwen3 Max