Qwen3.5 122B A10B heretic MTP NVFP4

Qwen 3

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

Released April 2026context N/A122B params

A large mixture-of-experts model that activates only 10 billion parameters per forward pass despite having 122 billion total, keeping inference costs manageable while drawing on a wide parameter pool. Published by OptimizeLLM under an Apache 2.0 license, it uses NVFP4 quantization and multi-token prediction, trading some raw precision for faster throughput. The 'heretic' tag suggests an experimental or community-modified variant, so behavior may differ from the standard Qwen3 release.

Capabilities

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

Creative Writing

Moderate

Multilingual

Qwen3.5 122B A10B heretic MTP NVFP4

Qwen 3

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

Released April 2026context N/A122B params

A large mixture-of-experts model that activates only 10 billion parameters per forward pass despite having 122 billion total, keeping inference costs manageable while drawing on a wide parameter pool. Published by OptimizeLLM under an Apache 2.0 license, it uses NVFP4 quantization and multi-token prediction, trading some raw precision for faster throughput. The 'heretic' tag suggests an experimental or community-modified variant, so behavior may differ from the standard Qwen3 release.

Capabilities

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

Creative Writing

Moderate

Multilingual

Glossary

Apache 2.0 LicenseA permissive open-source license that allows free use, modification, and distribution of software with minimal restrictions.Apache 2.0 LicenseAn open-source software license that allows free use, modification, and distribution of code with minimal restrictions.Community-ModifiedA model that has been altered or fine-tuned by users outside the original development team, often to change its behavior or capabilities.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.Multi-Token PredictionGenerating multiple future tokens in parallel instead of one at a time.Parameter PoolThe total set of learnable weights in a model; in sparse models, only a subset of this pool is activated for any given input.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.QuantizationReducing a model's numerical precision (e.g., from 16-bit to 4-bit) to shrink memory usage and speed up inference.ThroughputThe number of tokens a model can generate per second, measuring its processing speed.TokenA small unit of text (a word, subword, or punctuation mark) that a language model breaks input into for processing.Token PredictionThe core task of predicting what word or subword (token) should come next in a sequence based on previous text.

Capabilities

Capabilities

Use Case Fit

Glossary