ZDTaichu5.0-9B GGUF

GGUF files for ZDTaichu5.0-9B, an advanced multimodal foundation model developed by the Zi Dong Tai Chu team for visual understanding, spatial reasoning, agentic tool use, and embodied AI workloads. Built on a Qwen3.5-9B language backbone and C-RADIOv4-H vision encoder, the model supports a context length of up to 128K tokens (131,072).

For accelerated text generation via speculative decoding, this model can be paired with the companion draft model abenzerps/ZDTaichu5.0-9B-DSpark-GGUF.

Benchmarks

Benchmark results reported by TaichuAI for the original ZDTaichu5.0-9B foundation model. These figures are not measurements of this GGUF conversion.

GGUF files

Quantization	File	Size (GB)
Q8_0	ZDTaichu5.0-9B-Q8_0.gguf	9.53 GB
Q6_K	ZDTaichu5.0-9B-Q6_K.gguf	7.36 GB
Q5_K_M	ZDTaichu5.0-9B-Q5_K_M.gguf	6.47 GB
Q4_K_M	ZDTaichu5.0-9B-Q4_K_M.gguf	5.63 GB
Q4_0	ZDTaichu5.0-9B-Q4_0.gguf	5.31 GB
Q3_K_M	ZDTaichu5.0-9B-Q3_K_M.gguf	4.62 GB
IQ3_M	ZDTaichu5.0-9B-IQ3_M.gguf	4.42 GB
IQ3_XXS	ZDTaichu5.0-9B-IQ3_XXS.gguf	3.94 GB

IQ3_M and IQ3_XXS use importance-matrix quantization.

Multimodal projector

File	Size	Description
mmproj-ZDTaichu5.0-9B-BF16.gguf	1.65 GB	BF16 vision projector for runtimes with multimodal support

The projector is optional for text-only use. Use it with a current llama.cpp build that supports the model's multimodal path.

Chat template

The GGUF embeds the upstream chat template. An external copy is provided as chat_template.jinja for runtimes that require a separate template file.

Usage

For text generation with llama.cpp:

llama-cli \
  -m ZDTaichu5.0-9B-Q4_K_M.gguf \
  -c 8192 --jinja \
  --temp 0.7 --top-p 0.95 \
  -p "Explain why reproducible builds matter."

For multimodal vision understanding with llama.cpp:

llama-cli \
  -m ZDTaichu5.0-9B-Q4_K_M.gguf \
  --mmproj mmproj-ZDTaichu5.0-9B-BF16.gguf \
  --image path/to/image.jpg \
  -p "Describe what is shown in this image in detail."

For an OpenAI-compatible server:

llama-server \
  -m ZDTaichu5.0-9B-Q4_K_M.gguf \
  --mmproj mmproj-ZDTaichu5.0-9B-BF16.gguf \
  -c 8192 --jinja --host 0.0.0.0 --port 8080

Increase -c up to 131072 when sufficient memory is available. Tool-call behavior depends on the serving runtime and its parser integration; use the embedded template and verify tool calls in the target application.

Speculative Decoding (with DSpark)

Accelerate inference using the companion ZDTaichu5.0-9B-DSpark-GGUF draft model:

llama-cli \
  -m ZDTaichu5.0-9B-Q4_K_M.gguf \
  -md ZDTaichu5.0-9B-DSpark-Q4_0.gguf \
  -c 8192 --jinja \
  -p "Explain why reproducible builds matter."

Or with llama-server:

llama-server \
  -m ZDTaichu5.0-9B-Q4_K_M.gguf \
  -md ZDTaichu5.0-9B-DSpark-Q4_0.gguf \
  -c 8192 --jinja --host 0.0.0.0 --port 8080

Source and build

Source model: TaichuAI/ZDTaichu5.0-9B
Source revision: a22afd15a3f85659f103caa659ec4aa9500a998e
Conversion: ggml-org/llama.cpp commit 7ceed8737fdb4eb09b4760e77bd12d38012de5a8
License: Apache-2.0
Checksums: SHA256SUMS.txt

abenzerps/ZDTaichu5.0-9B-GGUF