# Qwen3.8-27B-DSpine-b16

> This checkpoint is a five-layer draft model for Qwen3.8-27B, with block size 16 and an 8,192-token sliding window.

- **Organisation:** tencent
- **Category:** Chat & Text Generation, Reasoning Models
- **Parameters:** 2B
- **Context length:** 262,144 tokens
- **Licence:** Apache 2.0 (Commercial use OK)
- **Released:** 2026-11-07
- **Hugging Face:** 5.6K downloads (30 days), 276 likes

## Estimated memory

| Quant | ≈ GB |
|---|---|
| Q2_K | 1.31 |
| Q3_K_M | 1.62 |
| Q4_K_M | 1.86 |
| Q5_K_M | 2.07 |
| Q6_K | 2.29 |
| Q8_0 | 2.76 |
| F16 | 4.62 |

## Runs with

- [MLX](https://github.com/ml-explore/mlx-lm) (community)

## About

This checkpoint is a five-layer draft model for Qwen3.8-27B, with block size 16 and an 8,192-token sliding window.

## Alternatives

- [openbmb/MiniCPM5-2B](https://selfhost.directory/models/openbmb/minicpm5-2b) (2.5B, Apache 2.0, ≈2.32 GB at Q4): We are releasing MiniCPM5-2B, the second model in the MiniCPM5 series, following MiniCPM5-1B. It is a dense 2B Transformer that scales up the same training recipe, built for on-device, local deployment, and resource-constrained scenarios, reaching 2B-class open-source SOTA.
- [Qwen/Qwen3-1.7B](https://selfhost.directory/models/qwen/qwen3-1.7b) (2B, Apache 2.0, ≈2.58 GB at Q4): Qwen3 is the latest generation of large language models in Qwen series, offering a comprehensive suite of dense and mixture-of-experts (MoE) models.
- [LiquidAI/LFM3-1.7B](https://selfhost.directory/models/liquidai/lfm3-1.7b) (1.7B, lfm1.0, ≈1.9 GB at Q4): LFM3-1.7B is a text model for local agents and specialized applications, based on our new tri-hybrid architecture.
- [LiquidAI/LFM3-1.7B-Base](https://selfhost.directory/models/liquidai/lfm3-1.7b-base) (1.7B, lfm1.0, ≈1.9 GB at Q4): LFM3-1.7B-Base is the pre-trained checkpoint behind LFM3-1.7B, a text model for local agents and specialized applications based on our new tri-hybrid architecture.
- [Qwen/Qwen3-1.7B-Base](https://selfhost.directory/models/qwen/qwen3-1.7b-base) (1.7B, Apache 2.0, ≈2.4 GB at Q4): Qwen3 is the latest generation of large language models in Qwen series, offering a comprehensive suite of dense and mixture-of-experts (MoE) models.
- [LiquidAI/LFM2.5-2.6B](https://selfhost.directory/models/liquidai/lfm2.5-2.6b) (2.7B, lfm1.0, ≈2.57 GB at Q4): LFM2.5-2.6B is part of LFM2.5, a family of hybrid models designed for on-device deployment. It builds on the LFM2 architecture with a 128K context window and agentic post-training.

More: https://selfhost.directory/models/tencent/qwen3.8-27b-dspine-b16/alternatives

## Runs well on

- [GeForce RTX 4060](https://selfhost.directory/models/tencent/qwen3.8-27b-dspine-b16/on/rtx-4060): F16, est. 37 tok/s
- [Apple M4 (16 GB)](https://selfhost.directory/models/tencent/qwen3.8-27b-dspine-b16/on/apple-m4-16gb): F16, est. 16 tok/s
- [Radeon RX 6700 XT](https://selfhost.directory/models/tencent/qwen3.8-27b-dspine-b16/on/rx-6700-xt): F16, est. 52 tok/s
- [GeForce RTX 3060 12GB](https://selfhost.directory/models/tencent/qwen3.8-27b-dspine-b16/on/rtx-3060): F16, est. 49 tok/s
- [Arc B580](https://selfhost.directory/models/tencent/qwen3.8-27b-dspine-b16/on/arc-b580): F16, est. 62 tok/s
- [Tesla T4](https://selfhost.directory/models/tencent/qwen3.8-27b-dspine-b16/on/tesla-t4): F16, est. 43 tok/s

## Similar size

- [LiquidAI/LFM3-1.7B](https://selfhost.directory/models/liquidai/lfm3-1.7b) (1.7B)
- [Qwen/Qwen2.5-1.5B-Instruct](https://selfhost.directory/models/qwen/qwen2.5-1.5b-instruct) (1.5B)
- [Qwen/Qwen3-1.7B](https://selfhost.directory/models/qwen/qwen3-1.7b) (2B)
- [openbmb/MiniCPM5-2B](https://selfhost.directory/models/openbmb/minicpm5-2b) (2.5B)
- [google/gemma-2-2b-it](https://selfhost.directory/models/google/gemma-2-2b-it) (2.6B)
- [HuggingFaceTB/SmolLM2-1.7B](https://selfhost.directory/models/huggingfacetb/smollm2-1.7b) (1.7B)

## FAQ

**How much VRAM does Qwen3.8-27B-DSpine-b16 need?**
About 1.9 GB at Q4_K_M with an 8K context. Q8_0 needs about 2.8 GB and full precision about 4.6 GB. Longer contexts need more memory for the KV cache.

**Can Qwen3.8-27B-DSpine-b16 run on the RTX 4060?**
Yes. It fits at FP16 / BF16 (about 4.6 GB), at an estimated 37 tokens/s. It fits comfortably on the Apple M4 (16 GB), Radeon RX 6700 XT, RTX 3060 12GB.

**Can I use Qwen3.8-27B-DSpine-b16 commercially?**
Yes. It is released under Apache 2.0. Permissive: use, modify and sell freely; keep the notice and licence text.

**What is the context length of Qwen3.8-27B-DSpine-b16?**
262,144 tokens (256K). The full window needs far more memory than an 8K context.

**How do I run Qwen3.8-27B-DSpine-b16 locally?**
With MLX: mlx_lm.server --model netprojects/Qwen3.8-27B-DSpine-b16-MLX --port 8080.

## Links

- huggingface: https://huggingface.co/tencent/Qwen3.8-27B-DSpine-b16
- homepage: https://dspine-project.github.io
- github: https://github.com/Tencent/workbuddy-bench
- paper: https://arxiv.org/abs/2609.36173

---
Source: https://selfhost.directory/models/tencent/qwen3.8-27b-dspine-b16