darkmaniac7/TokForge-AccelerationPack-Draft
Text Generation • Updated • 279
TokForge Draft Models
Small draft models TokForge pairs with a bigger chat model for speculative decoding. Research drafts have their own list.
Note Draft model for speculative decoding (KL-distilled). 4 GB RAM or more (TokForge helper module).
darkmaniac7/Qwen3-0.6B-lk-alpha-14b-paired-MNN
Text Generation • Updated • 79Note Draft model for speculative decoding (paired with Qwen3 14B). 4 GB RAM or more (TokForge helper module).
darkmaniac7/TokForge-AccelerationPack-Qwen35-Draft
Text Generation • Updated • 84Note Deprecated Qwen3.5 0.8B draft; the card says what to use instead.