Instructions to use jackasda211233/Qwen3.6-27B-AEON-RYS-15-20-GGUF with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Notebooks
Google Colab
Kaggle
Local Apps Settings

How to use jackasda211233/Qwen3.6-27B-AEON-RYS-15-20-GGUF with llama.cpp:

Install (macOS, Linux)

curl -LsSf https://llama.app/install.sh | sh
# Start a local OpenAI-compatible server with a web UI:
llama serve -hf jackasda211233/Qwen3.6-27B-AEON-RYS-15-20-GGUF:BF16
# Run inference directly in the terminal:
llama cli -hf jackasda211233/Qwen3.6-27B-AEON-RYS-15-20-GGUF:BF16

Install from WinGet (Windows)

winget install llama.cpp
# Start a local OpenAI-compatible server with a web UI:
llama serve -hf jackasda211233/Qwen3.6-27B-AEON-RYS-15-20-GGUF:BF16
# Run inference directly in the terminal:
llama cli -hf jackasda211233/Qwen3.6-27B-AEON-RYS-15-20-GGUF:BF16

Use pre-built binary

# Download pre-built binary from:
# https://github.com/ggerganov/llama.cpp/releases
# Start a local OpenAI-compatible server with a web UI:
./llama-server -hf jackasda211233/Qwen3.6-27B-AEON-RYS-15-20-GGUF:BF16
# Run inference directly in the terminal:
./llama-cli -hf jackasda211233/Qwen3.6-27B-AEON-RYS-15-20-GGUF:BF16

Build from source code

git clone https://github.com/ggerganov/llama.cpp.git
cd llama.cpp
cmake -B build
cmake --build build -j --target llama-server llama-cli
# Start a local OpenAI-compatible server with a web UI:
./build/bin/llama-server -hf jackasda211233/Qwen3.6-27B-AEON-RYS-15-20-GGUF:BF16
# Run inference directly in the terminal:
./build/bin/llama-cli -hf jackasda211233/Qwen3.6-27B-AEON-RYS-15-20-GGUF:BF16

Use Docker

docker model run hf.co/jackasda211233/Qwen3.6-27B-AEON-RYS-15-20-GGUF:BF16

LM Studio
Jan
Ollama
How to use jackasda211233/Qwen3.6-27B-AEON-RYS-15-20-GGUF with Ollama:
```
ollama run hf.co/jackasda211233/Qwen3.6-27B-AEON-RYS-15-20-GGUF:BF16
```
Unsloth Desktop

How to use jackasda211233/Qwen3.6-27B-AEON-RYS-15-20-GGUF with Pi:

Start the llama.cpp server

# Install llama.cpp:
brew install llama.cpp
# Start a local OpenAI-compatible server:
llama serve -hf jackasda211233/Qwen3.6-27B-AEON-RYS-15-20-GGUF:BF16

Configure the model in Pi

# Install Pi:
npm install -g @earendil-works/pi-coding-agent
# Add to ~/.pi/agent/models.json:
{
  "providers": {
    "llama-cpp": {
      "baseUrl": "http://localhost:8080/v1",
      "api": "openai-completions",
      "apiKey": "none",
      "models": [
        {
          "id": "jackasda211233/Qwen3.6-27B-AEON-RYS-15-20-GGUF:BF16"
        }
      ]
    }
  }
}

Run Pi

# Start Pi in your project directory:
pi

Docker Model Runner
How to use jackasda211233/Qwen3.6-27B-AEON-RYS-15-20-GGUF with Docker Model Runner:
```
docker model run hf.co/jackasda211233/Qwen3.6-27B-AEON-RYS-15-20-GGUF:BF16
```

Lemonade

How to use jackasda211233/Qwen3.6-27B-AEON-RYS-15-20-GGUF with Lemonade:

Pull the model

# Download Lemonade from https://lemonade-server.ai/
lemonade pull jackasda211233/Qwen3.6-27B-AEON-RYS-15-20-GGUF:BF16

Run and chat with the model

lemonade run user.Qwen3.6-27B-AEON-RYS-15-20-GGUF-BF16

List all available models

lemonade list

Hermes Agent

How to use jackasda211233/Qwen3.6-27B-AEON-RYS-15-20-GGUF with Hermes Agent:

Start the llama.cpp server

# Install llama.cpp:
brew install llama.cpp
# Start a local OpenAI-compatible server:
llama serve -hf jackasda211233/Qwen3.6-27B-AEON-RYS-15-20-GGUF:BF16

Configure Hermes

# Install Hermes:
curl -fsSL https://hermes-agent.nousresearch.com/install.sh | bash
hermes setup
# Point Hermes at the local server:
hermes config set model.provider custom
hermes config set model.base_url http://127.0.0.1:8080/v1
hermes config set model.default jackasda211233/Qwen3.6-27B-AEON-RYS-15-20-GGUF:BF16

Run Hermes

hermes

Atomic Chat

OpenClaw

How to use jackasda211233/Qwen3.6-27B-AEON-RYS-15-20-GGUF with OpenClaw:

Start the llama.cpp server

# Install llama.cpp:
brew install llama.cpp
# Start a local OpenAI-compatible server:
llama serve -hf jackasda211233/Qwen3.6-27B-AEON-RYS-15-20-GGUF:BF16

Configure OpenClaw

# Install OpenClaw:
npm install -g openclaw@latest
# Register the local server and set it as the default model:
openclaw onboard --non-interactive --mode local \
  --auth-choice custom-api-key \
  --custom-base-url http://127.0.0.1:8080/v1 \
  --custom-model-id "jackasda211233/Qwen3.6-27B-AEON-RYS-15-20-GGUF:BF16" \
  --custom-provider-id llama-cpp \
  --custom-compatibility openai \
  --custom-text-input \
  --accept-risk \
  --skip-health

Run OpenClaw

openclaw agent --local --agent main --message "Hello from Hugging Face"

jackasda211233 commited on May 4

Commit

f131287

verified ·

1 Parent(s): d153963

Move benchmarks to top of model card

Browse files

Files changed (1) hide show

README.md +45 -49

README.md CHANGED Viewed

@@ -23,65 +23,61 @@ base_model:
 # Qwen3.6-27B-AEON-RYS-MaxThinkCoder — Speedup Research + Experimental MTP GGUF
-This page now includes two GGUF paths:
-- **Recommended default:** the original non-MTP custom IQ4_NL build. It is still the standard release because it is faster in our tests, more accurate in practical evals, and less prone to long-output repetition.
-- **Experimental MTP build:** a true MTP-capable GGUF for people who want to test `ik_llama.cpp` multi-token prediction on this RYS model. It is interesting technically, but it did **not** beat the non-MTP file in our speed or practical-quality tests.
-The word “speedup” here refers to the custom `ik_llama` runtime work and the MTP speedup experiment. It is **not** a claim that the MTP file is faster than the recommended non-MTP model. The author recommendation is still: use the normal non-MTP GGUF unless you specifically want to experiment with MTP.
-The `SpeedBoosted` wording in the experimental MTP filename refers to this release/fork context: the custom `ik_llama` runtime is substantially faster than the patched/base runtime path on this model family. It does not mean MTP beat the recommended non-MTP IQ4_NL file in our tests.
-Hyper-focused Q4NL RYS release for:
-- programming
-- technical reasoning
-- academic-style writing
-This release is built from:
-- AEON source model:
-  `https://huggingface.co/AEON-7/Qwen3.6-27B-AEON-Ultimate-Uncensored`
-Use this model with:
-- custom `ik-llama` fork, specialized and tuned for this exact model:
-  `https://github.com/noonr48/qwen36-aeon-ik-llama`
-- the current fork supports both the recommended non-MTP GGUF and the experimental MTP GGUF; the non-MTP path was rechecked after the MTP work and still loads/runs normally
-RYS build guide:
-- practical guide / write-up:
-  `https://github.com/noonr48/qwen36-aeon-ik-llama/tree/main/docs/rys-layer-duplication-guide`
-- PDF snapshot:
-  `https://github.com/noonr48/qwen36-aeon-ik-llama/blob/main/docs/rys-layer-duplication-guide/rys_layer_duplication_guide.pdf`
-Side note (tool calling):
-some prompts can trigger repeated *identical* tool calls in one assistant turn (especially when the tool result is empty / slow).
-Update to the latest `ik-llama` fork version: it now deduplicates identical `tool_calls` server-side.
-## At a glance
-- released Q4_NL GGUF:
-  [`Qwen3.6-27B-AEON-RYS-MaxThinkCoder-IQ4_NL-ik-llama-custom-mixed.gguf`](https://huggingface.co/jackasda211233/Qwen3.6-27B-AEON-RYS-15-20-GGUF/blob/main/Qwen3.6-27B-AEON-RYS-MaxThinkCoder-IQ4_NL-ik-llama-custom-mixed.gguf)
-- experimental MTP IQ4_NL GGUF:
-  [`Qwen3.6-27B-AEON-RYS-MaxThinkCoder-SpeedBoosted-IQ4_NL-MTP-Experimental.gguf`](https://huggingface.co/jackasda211233/Qwen3.6-27B-AEON-RYS-15-20-GGUF/blob/main/Qwen3.6-27B-AEON-RYS-MaxThinkCoder-SpeedBoosted-IQ4_NL-MTP-Experimental.gguf)
-- BF16 GGUF reference:
-  [`Qwen3.6-27B-AEON-RYS-MaxThinkCoder-BF16.gguf`](https://huggingface.co/jackasda211233/Qwen3.6-27B-AEON-RYS-15-20-GGUF/blob/main/Qwen3.6-27B-AEON-RYS-MaxThinkCoder-BF16.gguf)
-- HF-format BF16 safetensors:
-  [`bf16-safetensors/`](https://huggingface.co/jackasda211233/Qwen3.6-27B-AEON-RYS-15-20-GGUF/tree/main/bf16-safetensors)
-- RYS build guide:
-  [https://github.com/noonr48/qwen36-aeon-ik-llama/tree/main/docs/rys-layer-duplication-guide](https://github.com/noonr48/qwen36-aeon-ik-llama/tree/main/docs/rys-layer-duplication-guide)
-- intended runtime:
-  custom `ik-llama`
-- compression:
-  `54G` BF16 -> `16G` IQ4_NL
-- mixed validation snapshot:
-  `0.7299` BF16 -> `0.7244` IQ4_NL
-- overall performance change:
-  `-0.0055` absolute, about `-0.75%` relative
-- public codename:
-  `MaxThinkCoder`
-- project focus:
-  get as much capability as possible out of a Q4-class RYS model for hard English-first work
-- not a generic chat release
-- not a stock `llama.cpp` release
 ## Experimental MTP GGUF

 # Qwen3.6-27B-AEON-RYS-MaxThinkCoder — Speedup Research + Experimental MTP GGUF
+## Benchmark + Key Details
+Use the normal non-MTP `IQ4_NL-ik-llama-custom-mixed` GGUF for actual work. The MTP GGUF is published for runtime research; it loads and has the expected MTP tensors, but it did **not** beat the normal file in our speed or practical-quality tests.
+### Download Targets
+- **Recommended default GGUF:**
+  [`Qwen3.6-27B-AEON-RYS-MaxThinkCoder-IQ4_NL-ik-llama-custom-mixed.gguf`](https://huggingface.co/jackasda211233/Qwen3.6-27B-AEON-RYS-15-20-GGUF/blob/main/Qwen3.6-27B-AEON-RYS-MaxThinkCoder-IQ4_NL-ik-llama-custom-mixed.gguf)
+- **Experimental MTP GGUF:**
+  [`Qwen3.6-27B-AEON-RYS-MaxThinkCoder-SpeedBoosted-IQ4_NL-MTP-Experimental.gguf`](https://huggingface.co/jackasda211233/Qwen3.6-27B-AEON-RYS-15-20-GGUF/blob/main/Qwen3.6-27B-AEON-RYS-MaxThinkCoder-SpeedBoosted-IQ4_NL-MTP-Experimental.gguf)
+- **BF16 GGUF reference:**
+  [`Qwen3.6-27B-AEON-RYS-MaxThinkCoder-BF16.gguf`](https://huggingface.co/jackasda211233/Qwen3.6-27B-AEON-RYS-15-20-GGUF/blob/main/Qwen3.6-27B-AEON-RYS-MaxThinkCoder-BF16.gguf)
+- **HF-format BF16 safetensors:**
+  [`bf16-safetensors/`](https://huggingface.co/jackasda211233/Qwen3.6-27B-AEON-RYS-15-20-GGUF/tree/main/bf16-safetensors)
+### Quality / Compression
+The released Q4 model is about `70%` smaller on disk:
+`54G` BF16 -> `16G` IQ4_NL.
+| probe | BF16 | released IQ4_NL |
+|---|---:|---:|
+| mixed 4-probe mean | `0.7299` | `0.7244` |
+| `math_16` | `0.8421` | `0.7897` |
+| `eq_16` | `0.7123` | `0.7111` |
+| `math_4` | `0.4851` | `0.5170` |
+| `gsm8k_5` | `0.8800` | `0.8800` |
+Net change: `-0.0055` absolute, about `-0.75%` relative on the mixed snapshot.
+### Speed / MTP
+| check | setup | decode speed |
+|---|---|---:|
+| recommended custom `ik-llama` deployment | 6x RTX 5060 Ti, `ctx=409600`, `np=2`, `f32/f32` KV | `39.37 tok/s` |
+| patched upstream-style comparison path | 6x RTX 5060 Ti, `ctx=4096`, `np=1`, `f16` KV | `22.51 tok/s` |
+| no-MTP short generation | 3x RTX 3090, graph split, 192-token check | `48.68 tok/s` |
+| naive MTP draft-1 short generation | 3x RTX 3090, same 192-token check | `38.16 tok/s` |
+| adaptive MTP short generation | 3x RTX 3090, same 192-token check | `45.36 tok/s` |
+| adaptive MTP longer generation | 3x RTX 3090, 768-token check | `46.95 tok/s` vs `48.71 tok/s` no-MTP |
+The `SpeedBoosted` wording in the MTP filename refers to the custom `ik_llama` runtime/release context. It does **not** mean the MTP file beat the recommended non-MTP IQ4_NL file.
+### Model Details
+- Public codename: `MaxThinkCoder`
+- Source model: [`AEON-7/Qwen3.6-27B-AEON-Ultimate-Uncensored`](https://huggingface.co/AEON-7/Qwen3.6-27B-AEON-Ultimate-Uncensored)
+- RYS branch: AEON-derived `blocks:15,20`
+- Focus: coding, technical reasoning, academic-style writing, and long-context English work
+- Intended runtime: custom [`ik-llama`](https://github.com/noonr48/qwen36-aeon-ik-llama)
+- RYS build guide: [`docs/rys-layer-duplication-guide`](https://github.com/noonr48/qwen36-aeon-ik-llama/tree/main/docs/rys-layer-duplication-guide)
+- Not a generic safe-default chat model
+- Not a stock `llama.cpp` release
+Side note: update to the latest `ik-llama` fork if you use tools. It deduplicates repeated identical `tool_calls` server-side.
 ## Experimental MTP GGUF