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.
Qwen3.8-27B-DSpine-b16
tencent·Base model·released 7 Nov 2026
mlx_lm.server --model netprojects/Qwen3.8-27B-DSpine-b16-MLX --port 8080This checkpoint is a five-layer draft model for Qwen3.8-27B, with block size 16 and an 8,192-token sliding window.
How much VRAM does Qwen3.8-27B-DSpine-b16 need?
At Q4_K_M Qwen3.8-27B-DSpine-b16 needs about 1.9 GB with an 8K context (1.9 GB at Q4, 4.6 GB at full precision). Pick your GPU to see what fits and how fast it runs.
Needs: ≈1.9 GB · Q4_K_M · 8K context — Pick a GPU to check the fit.
Advanced: context 8K · KV cache 16-bit
Smallest GPU for each quant and context (KV cache 16-bit)
| Quantisation | 8K | 16K | 32K | 64K | 128K | 256K |
|---|---|---|---|---|---|---|
| Q3_K_M | ||||||
| Q4_K_M | ||||||
| Q5_K_M | ||||||
| Q6_K | ||||||
| Q8_0 | ||||||
| FP16 |
≤16 GB card24–32 GB48–96 GBmore
| Quantisation | Total |
|---|---|
| Q2_KHeavy lossSmallest, noticeably worse quality | ≈1.3 GB |
| Q3_K_MNoticeable lossSmall, some quality loss | ≈1.6 GB |
| Q4_K_MrecommendedSmall lossRecommended default for local use | ≈1.9 GB |
| Q5_K_MNear-losslessNear-lossless for most tasks | ≈2.1 GB |
| Q6_KNear-losslessPractically lossless | ≈2.3 GB |
| Q8_0LosslessLossless in practice, twice the size of Q4 | ≈2.8 GB |
| FP16 / BF16LosslessFull precision, as released | ≈4.6 GB |
Estimates for 8K tokens of context: weights + KV cache + runtime overhead. Speed assumes memory-bound decoding at your GPU’s bandwidth. Pick a GPU above and the fit column checks every row against it. Weights marked “file” use the real GGUF download size; “est.” rows are computed from the parameter count.
How to pick a quant
Take the biggest quant that fits with room left for context. Q4_K_M is the usual sweet spot. Below Q4 quality drops fast, above Q6 you mostly pay memory for nothing. Big models (70B+, large MoE) tolerate low quants better than small ones.
- LosslessF16, Q8_0
- Near-losslessQ6_K, Q5_K_M
- Small lossQ5_K_S, Q4_K_M, IQ4_NL, IQ4_XS, Q4_K_S
- Noticeable lossQ4_0, Q3_K_L, Q3_K_M, IQ3_M
- Heavy lossIQ3_XXS, Q3_K_S, IQ2_M, Q2_K
Typical quality loss, not measured per model. Tiers come from KL divergence against FP16 on Llama 3 8B: llama.cpp perplexity README: LLaMA 3 8B scoreboard (KLD and ΔPPL vs FP16, Wikitext-2); llama.cpp quantize tool: ΔPPL per quant type @ Llama-3-8B
How much context fits?
At 8K context, Q4_K_M needs ≈1.9 GB(1.2 GB weights + 0.2 GB KV cache + 0.6 GB overhead). Pick a GPU above to see how much context fits.
Weights + overheadKV cache
Can the RTX 4060 run Qwen3.8-27B-DSpine-b16? →
Speed vs quality per quant
Smaller quants run faster and fit in less memory, but lose quality.
Fits in VRAMTight fitCPU offload✕ won't runQuality is typical KL divergence vs FP16, not measured for this model.
Real-world speed of Qwen3.8-27B-DSpine-b16
Tokens per second measured by people running it at home. A GPU shows a median once 3 reports agree on the same setup. Reports far from our estimate wait for a review. Measure yours with llama-bench →
No medians yet. Be one of the first 3 to report a setup.
All reports
Run Qwen3.8-27B-DSpine-b16 locally
These commands load the original weights from Hugging Face.
pip install mlx-lm
mlx_lm.generate --model netprojects/Qwen3.8-27B-DSpine-b16-MLX --prompt "Hello"Use `mlx_lm.chat` with the same --model for an interactive chat.
mlx_lm.server --model netprojects/Qwen3.8-27B-DSpine-b16-MLX --port 8080API on http://localhost:8080/v1.
Quantised versions and downloads
| Repository | Format | Quant | Size |
|---|---|---|---|
| aj9o9/Qwen3.8-27B-DSpine-b16-GGUF | GGUF | Q4_K_M | 1.3 GB |
| aj9o9/Qwen3.8-27B-DSpine-b16-GGUF | GGUF | Q8_0 | 2.2 GB |
| aj9o9/Qwen3.8-27B-DSpine-b16-GGUF | GGUF | BF16 | 4.1 GB |
| netprojects/Qwen3.8-27B-DSpine-b16-MLX | MLX | — | — |
About 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.
Full model card on Hugging Face →Alternatives to Qwen3.8-27B-DSpine-b16
All alternatives →Qwen3 is the latest generation of large language models in Qwen series, offering a comprehensive suite of dense and mixture-of-experts (MoE) models.
LFM3-1.7B is a text model for local agents and specialized applications, based on our new tri-hybrid architecture.
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.
Qwen3 is the latest generation of large language models in Qwen series, offering a comprehensive suite of dense and mixture-of-experts (MoE) models.
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.
Qwen3.8-27B-DSpine-b16 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.
Rate this model
Data from Hugging Face, the model card and public leaderboards. Memory and speed figures are estimates; real numbers depend on the runtime, context length and drivers. Always check the licence yourself before commercial use.
0 comments