Qwen3.8-27B-OBLITERATED-V3-Sharp-oQ4e-mtp

MLX oQ4e quantization of OBLITERATUS/Qwen3.8-27B-OBLITERATED (V3 – Deep Liberation).

Quantized with oMLX v0.6.3rc3.

The model includes the MTP (Multi-Token Prediction) weights for speculative decoding with compatible runtimes.

The original vision components have been removed to reduce memory usage. This release is therefore text-only.

It also includes the Qwen Sharp Chat Template, optimized for more concise and token-efficient Qwen 3.8 responses.

Quantization details

• Base model: OBLITERATUS/Qwen3.8-27B-OBLITERATED (V3)
• Original base: Qwen/Qwen3.8-27B
• Quantization: oQ4e
• Quantizer: oMLX 0.6.3rc3
• Format: MLX Safetensors
• MTP: Included
• Vision: Removed
• Chat template: Qwen Sharp