Qwen3.5 397B A17B MXFP4

Qwen 3

Open WeightModel weights are publicly available — can be downloaded and self-hosted

Released March 2026context N/A

A large mixture-of-experts model that activates roughly 17 billion parameters per forward pass despite having 397 billion total parameters, keeping inference costs manageable. It handles both text and image inputs, making it capable of multimodal reasoning tasks. The MXFP4 quantization reduces memory footprint, though this precision trade-off may affect output fidelity compared to higher-precision variants.

Capabilities

Capability scores are AI-generated based on model documentation, benchmarks, and technical specifications. Learn more

Instruction Following

Moderate

Tool Use

Moderate

Creative Writing

Use Case Fit

Fit scores are AI-generated based on model capabilities, intended use, and technical specifications. Learn more

Qwen3.5 397B A17B MXFP4

Qwen 3

Open WeightModel weights are publicly available — can be downloaded and self-hosted

Released March 2026context N/A

A large mixture-of-experts model that activates roughly 17 billion parameters per forward pass despite having 397 billion total parameters, keeping inference costs manageable. It handles both text and image inputs, making it capable of multimodal reasoning tasks. The MXFP4 quantization reduces memory footprint, though this precision trade-off may affect output fidelity compared to higher-precision variants.

Capabilities

Capability scores are AI-generated based on model documentation, benchmarks, and technical specifications. Learn more

Instruction Following

Moderate

Tool Use

Moderate

Creative Writing

Use Case Fit

Fit scores are AI-generated based on model capabilities, intended use, and technical specifications. Learn more

Glossary

FidelityThe degree to which a quantized or compressed model preserves the quality and accuracy of the original full-precision model.Forward PassA single computation cycle where input data flows through the model's layers to produce an output prediction.InferenceThe process of running a trained model to generate predictions or outputs from new inputs.MXFP4A low-precision floating-point format (4-bit) designed for efficient neural network computation while maintaining reasonable accuracy.MXFP4 QuantizationA low-precision numerical format that represents model weights and activations using only 4 bits, significantly reducing memory usage and inference speed at the cost of some accuracy.Memory FootprintThe amount of RAM or storage space a model requires to run, which is critical for deployment on resource-constrained devices.MultimodalA model that can process and understand multiple types of input, such as both text and images.Multimodal ReasoningThe ability to solve problems by integrating information from multiple input types like images and text.ParametersThe learned numerical values in a model — more parameters generally means more capacity but higher compute cost.PrecisionThe level of numerical detail a model uses to represent its internal values; higher precision means more accurate calculations but requires more memory.Precision Trade-offThe balance between reducing model size through lower numerical precision and maintaining accuracy—lower precision saves memory but may slightly reduce performance.QuantizationReducing a model's numerical precision (e.g., from 16-bit to 4-bit) to shrink memory usage and speed up inference.ReasoningThe model's ability to work through multi-step logical problems and provide justified answers rather than just pattern-matching.Reasoning TasksProblems that require a model to think through multiple steps logically to arrive at an answer, rather than just pattern-matching.