Skip to content
C-Kernel-Engine docs Antshiv Open Source

C-KERNEL-ENGINE · GUIDE

Qwen3.8 27B dense

Capability matrix

  • Compute and format Q4_K_M · Ryzen 9 9950X3D Zen 5 artifact runs.
  • Local entry point Text REPL.
  • Server Recorded HTTP pass. Identified 32K dense bundle.
  • Harness Qwen Code. Repeated read, edit and test tasks. Recheck each new quantization or context build.

Choose the download

Publisher model card · CKE input files

Publisher task: image text to text. CKE example: Q4_K_M GGUF. Use this artifact identifier:

hf://ggml-org/Qwen3.8-27B-GGUF/Qwen3.8-27B-Q4_K_M.gguf.

Get started

High-memory artifact; use a fresh run directory with --force-convert --force-compile and follow the v8 runbook's tested command and evidence boundary

Examplebash
version/v8/scripts/cks-v8-run run \
  hf://ggml-org/Qwen3.8-27B-GGUF/Qwen3.8-27B-Q4_K_M.gguf \
  --context-len 16384 \
  --force-convert --force-compile \
  --chat-template auto \
  --prompt 'Give me a detailed example of C, Python and SQL code.' \
  --max-tokens 256 \
  --temperature 0.0

What has actually run?

Earlier artifact: 4,000/4,000 full-vocabulary rows bit-exact to llama.cpp; six complete long-context artifact runs. 4,096-token exact trajectory; 131,072-token execution; 262,144-token planner.

CPU evidence. Ryzen 9 9950X3D Zen 5 artifact runs.

Performance. Earlier artifact on Ryzen 9 9950X3D, six 131,072-token runs: 32.37–32.71 tokens/s prefill and 1.36–1.41 tokens/s decode. Recheck the artifact revision before comparing. ARM throughput for this exact artifact is unverified here. Benchmark evidence · 32K to 128K results

Can I fine-tune it?

Full-checkpoint training, LoRA and other PEFT methods: no verified CKE recipe for this exact checkpoint in the reviewed evidence. For a trainable starter, use the FP32 and adapter recipes.

Start an interactive session

Run from the CKE checkout. Type a question, then /exit to finish.

Examplebash
version/v8/scripts/cks-v8-run run \
  hf://ggml-org/Qwen3.8-27B-GGUF/Qwen3.8-27B-Q4_K_M.gguf \
  --chat-template auto --max-tokens 128 --temperature 0

Configure the tested server and client.

Try the HTTP server

For a short plain-text request, install the server dependencies and use a separate build directory. This command is a starting configuration. The capability card above records the checks completed for this artifact. Larger tool prompts need the context and serving variant specified by their harness.

Examplebash
version/v8/scripts/cks-v8-run serve \
  hf://ggml-org/Qwen3.8-27B-GGUF/Qwen3.8-27B-Q4_K_M.gguf \
  --run "$HOME/cke-runs/qwen3-8-27b-dense-chat" \
  --context-len 1024 --max-tokens 64 --model-name cke-local \
  --host 127.0.0.1 --port 8080

Check the model ID and send one real request.