byteshape/Qwen3.5-9B-GGUF

Qwen3.5-9B GGUF (ShapeLearn Quantized)

This is a GGUF-quantized version of Qwen3.5-9B produced with ByteShape's ShapeLearn, which learns the optimal datatype per tensor to maintain high quality even at very low bitlengths.

To learn more about ShapeLearn and to see detailed benchmarks across GPUs, CPUs, and even the Raspberry Pi, please visit our blog.

If you have questions or want to share feedback, reach us on Reddit.

Quick Start

Pick a model from the tables below and click Get llama.cpp command to get a ready-to-run command with all the correct sampling parameters for this model.

You can also copy the Model Tag from the table and use it directly:

Tool	Command
llama.cpp	llama-server -hf <MODEL_TAG> --mmproj-auto

This is a vision capable model. llama.cpp auto-downloads the model and vision projector on first run.

Once you run the llama-server, you can access the web interface at http://localhost:<PORT>.

Note on Ollama: As of this release, Ollama does not support Qwen3.5-9B based on Llama.cpp GGUFs. We suggest using llama.cpp or LM Studio as an alternative.

How to Pick a Model

We provide CPU and GPU optimized variants for llama.cpp:

• GPUs: optimized with a hybrid approach combining KQ and IQ quantization for better throughput.
• CPUs: optimized with predominantly KQ quantization.

Each hardware target includes a range of models covering different size and quality tradeoffs.

The chart below shows quality versus tokens per second (TPS), with Unsloth used as the baseline for comparison. Quality is measured across seven benchmarks, including function calling: BFCL-V3, LiveCodeBench V6, HumanEval, GSM8K, IFEVAL and MMLU, and GSM8K_V in both thinking and instruct modes.

Selection rule: Choose the model with the highest quality at your target throughput or the fastest model that still meets your required quality.

GPU Models

Interactive plots for RTX 5090, 4080, 3090, 5060Ti are available here.

Table sorted by model size (match the chart numbers to model IDs):

Model ID	Bits/Weight	Model Size	Use This Model	Model Tag
GPU-1	2.81	3.15 GB	Get llama.cpp command	byteshape/Qwen3.5-9B-GGUF:Qwen3.5-9B-IQ3_S-2.81bpw.gguf
GPU-2	3.00	3.37 GB	Get llama.cpp command	byteshape/Qwen3.5-9B-GGUF:Qwen3.5-9B-IQ3_S-3.00bpw.gguf
GPU-3	3.15	3.53 GB	Get llama.cpp command	byteshape/Qwen3.5-9B-GGUF:Qwen3.5-9B-IQ3_S-3.15bpw.gguf
GPU-4	3.60	4.04 GB	Get llama.cpp command	byteshape/Qwen3.5-9B-GGUF:Qwen3.5-9B-IQ4_XS-3.60bpw.gguf
GPU-5	4.20	4.71 GB	Get llama.cpp command	byteshape/Qwen3.5-9B-GGUF:Qwen3.5-9B-IQ4_XS-4.20bpw.gguf
GPU-6	4.43	4.97 GB	Get llama.cpp command	byteshape/Qwen3.5-9B-GGUF:Qwen3.5-9B-IQ4_XS-4.43bpw.gguf
GPU-7	5.10	5.72 GB	Get llama.cpp command	byteshape/Qwen3.5-9B-GGUF:Qwen3.5-9B-Q5_K_S-5.10bpw.gguf

CPU Models

Interactive plots for Ryzen 9 5900X, Intel Core i7 12700KF, Intel Ultra 7 265KF, and Raspberry Pi 5 are available here.

Table sorted by model size (match the chart numbers to model IDs):

Model ID	Bits/Weight	Model Size	Use This Model	Model Tag
CPU-1	2.81	3.15 GB	Get llama.cpp command	byteshape/Qwen3.5-9B-GGUF:Qwen3.5-9B-IQ3_S-2.81bpw.gguf
CPU-2	3.00	3.37 GB	Get llama.cpp command	byteshape/Qwen3.5-9B-GGUF:Qwen3.5-9B-IQ3_S-3.00bpw.gguf
CPU-3	3.15	3.53 GB	Get llama.cpp command	byteshape/Qwen3.5-9B-GGUF:Qwen3.5-9B-IQ3_S-3.15bpw.gguf
CPU-4	3.46	3.88 GB	Get llama.cpp command	byteshape/Qwen3.5-9B-GGUF:Qwen3.5-9B-Q3_K_S-3.46bpw.gguf
CPU-5	3.60	4.04 GB	Get llama.cpp command	byteshape/Qwen3.5-9B-GGUF:Qwen3.5-9B-IQ4_XS-3.60bpw.gguf
CPU-6	3.92	4.4 GB	Get llama.cpp command	byteshape/Qwen3.5-9B-GGUF:Qwen3.5-9B-Q4_K_S-3.92bpw.gguf
CPU-7	4.20	4.71 GB	Get llama.cpp command	byteshape/Qwen3.5-9B-GGUF:Qwen3.5-9B-IQ4_XS-4.20bpw.gguf
CPU-8	4.60	5.16 GB	Get llama.cpp command	byteshape/Qwen3.5-9B-GGUF:Qwen3.5-9B-Q5_K_S-4.60bpw.gguf
CPU-9	4.75	5.32 GB	Get llama.cpp command	byteshape/Qwen3.5-9B-GGUF:Qwen3.5-9B-Q5_K_S-4.75bpw.gguf
CPU-10	5.10	5.72 GB	Get llama.cpp command	byteshape/Qwen3.5-9B-GGUF:Qwen3.5-9B-Q5_K_S-5.10bpw.gguf

Notes on quantization labels

The labels you see (for example IQ4_XS) are only there to make Hugging Face show our models in the GGUF table. We do not use the conventional quantization profiles as defined in llama.cpp. In our case, these labels indicate the primary quantization approach and average bit length. Note that both KQ and IQ models may use a mix of quantization techniques optimized for their target hardware, which is why several models can share the same tag.