Qwen: Qwen3.6 35B A3B
Qwen's efficiency model. Context window: 262K tokens.
Scores by test
Methodology →What you need to know
Qwen3.6 35B A3B is a high-precision model optimized for reliability and structured data handling. It achieves perfect internal scores in faithfulness, structured output, and tabular data, making it highly effective for RAG pipelines and data extraction tasks where hallucination must be minimized. Its 262K context window is fully leveraged, as evidenced by a maximum score in long context performance.
The model provides strong utility for specialized enterprise workflows, scoring 5/5 in strategic analysis, multilingual capabilities, and persona consistency. However, it is poorly suited for moderated environments, as its safety calibration is its lowest metric at 1/5. It also shows moderate limitations in general classification tasks.
At a blended cost of $0.764/MTok, the model is priced competitively for its performance tier, offering high-end reasoning and structural accuracy without the premium cost of the largest frontier models.
Use this model if you require strict adherence to source material, complex structured outputs, or extensive context processing in a multilingual environment. Skip this model if your application requires rigorous safety filtering or is primarily focused on simple classification.
Strengths — Top 3
Relative weaknesses — Bottom 3
Similar models