Schestex/ThinkingCap-Qwen3.8-27B-NInfer

ThinkingCap-Qwen3.8-27B NInfer

Native NInfer artifact of ThinkingCap-Qwen3.8-27B, optimized for local inference on NVIDIA Blackwell GPUs.

This repository contains a .ninfer artifact only. It is not a Transformers checkpoint, GGUF model or Safetensors distribution.

---

Features

• Native NInfer artifact
• Qwen3.8-27B ThinkingCap
• MTP (Multi-Token Prediction) support
• Long-context capable
• Optimized for NVIDIA Blackwell GPUs
• Compatible with the current NInfer runtime

---

Download

Download the complete repository:

hf download Schestex/ThinkingCap-Qwen3.8-27B-NInfer \
    --local-dir ./ThinkingCap-Qwen3.8-27B-NInfer

Or download only the artifact:

hf download Schestex/ThinkingCap-Qwen3.8-27B-NInfer \
    thinkingcap_qwen3_8_27b.ninfer \
    --local-dir models

---

Example

./build/apps/ninfer \
    models/thinkingcap_qwen3_8_27b.ninfer \
    --prompt "Explain speculative decoding." \
    --max-context 32768 \
    --max-new 1024 \
    --spec mtp \
    --draft-tokens 3

---

Requirements

• Recent NInfer runtime
• CUDA 13+
• NVIDIA Blackwell GPU recommended

---

About ThinkingCap

ThinkingCap is a reasoning-focused variant of Qwen3.8-27B designed for high-quality assistant workloads.

This repository provides the model packaged in the native NInfer artifact format for efficient local inference.

---

Credits

Base model:

• bottlecapai/ThinkingCap-Qwen3.8-27B

Runtime and artifact format:

• NInfer

---

License

Please follow the license terms of the original ThinkingCap model.

This repository only distributes the converted NInfer artifact.