Savyasaachin/Qwen3-Shining-Lucy-CODER-3.4B-Brainstorm20x-e32-Q8_0-GGUF
This model was converted to GGUF format from DavidAU/Qwen3-Shining-Lucy-CODER-3.4B-Brainstorm20x-e32 using llama.cpp via the ggml.ai's GGUF-my-repo space.
Refer to the original model card for more details on the model.
Use with llama.cpp
Install llama.cpp through brew (works on Mac and Linux)
brew install llama.cpp
Invoke the llama.cpp server or the CLI.
CLI:
llama-cli --hf-repo Savyasaachin/Qwen3-Shining-Lucy-CODER-3.4B-Brainstorm20x-e32-Q8_0-GGUF --hf-file qwen3-shining-lucy-coder-3.4b-brainstorm20x-e32-q8_0.gguf -p "The meaning to life and the universe is"
Server:
llama-server --hf-repo Savyasaachin/Qwen3-Shining-Lucy-CODER-3.4B-Brainstorm20x-e32-Q8_0-GGUF --hf-file qwen3-shining-lucy-coder-3.4b-brainstorm20x-e32-q8_0.gguf -c 2048
Note: You can also use this checkpoint directly through the usage steps listed in the Llama.cpp repo as well.
From the Original Creator:
This model contains - "ValiantLabs/Qwen3-1.7B-ShiningValiant3" FUSED with "Menlo/Lucy-128k" (instruct models):
- https://huggingface.co/ValiantLabs/Qwen3-1.7B-ShiningValiant3
- https://huggingface.co/Menlo/Lucy-128k
Fused formula by DavidAU, which co-joins the models in 4 blocks, in a roughly 75%/75% merge resulting in a model of 42 layers, 464 tensors creating a 2.4B model from the TWO 1.7B models THEN Brainstorm 20x is added making the model 3.4B, 61 layers, and 673 tensors.
This method takes the best of both models without compromise or "averaging" / "cancelling out" parts of either model, increasing model net performance and retaining model knowledge.
The Brainstorm adapter improves code generation, and unique code solving abilities - in the case of this model combo, it provides out of the box "code generation".
This version (e32) uses FLOAT 32 source, to take advantage of ShiningValiant3 being in float32. This model will be a lot stronger than its bfloat16 version counterpart as a result.
Quants built from this source will also be stronger.
The 2.4B model is also a reasoning model, with a lot shorter "reasoning" blocks - part of the FUSED merge effect.
Settings information on the 2.4B fused model below, followed by info from each model (model card), Brainstorm 20x adapter, and then a complete help section for running LLM / AI models.
This model requires:
- Jinja (embedded) or CHATML template
- Max context of 40k.
Settings used for testing (suggested):
- Temp .3 to .7
- Rep pen 1.05 to 1.1
- Topp .8 , minp .05
- Topk 20
- No system prompt.
- Min context window of 8k to 16k suggested.
- Suggest 2-4 generations due to Brainstorm adapter and unique code generation each time.
BEST settings for coding:
- Temp .8, rep pen 1.05 OR rep pen 1.1
- topk 20, top p .95, minp 0
- Context window 16k.
- Suggest 2-4 generations due to Brainstorm adapter and unique code generation each time.
- Then after 2-4 generations, raise temp to .9 or lower to .5 -> regen 2-4 times.
FOR CODING:
Higher temps: .6 to .9 (even over 1) work better for more complex coding / especially with more restrictions.
This model will respond well to both detailed instructions and step by step refinement and additions to code.
As this is an instruct model, it will also benefit from a detailed system prompt too.
For simpler coding problems, lower quants will work well; but for complex/multi-step problem solving suggest Q6 or Q8.
QUANTS:
Special Thanks to Team Mradermacher for the quants:
GGUF:
https://huggingface.co/mradermacher/Qwen3-Shining-Lucy-CODER-3.4B-Brainstorm20x-e32-GGUF
GGUF-IMATRIX:
https://huggingface.co/mradermacher/Qwen3-Shining-Lucy-CODER-3.4B-Brainstorm20x-e32-i1-GGUF
Model #1 - Shining Valiant 1.7B
Support our open-source dataset and model releases!
Shining Valiant 3: Qwen3-1.7B, Qwen3-8B
Shining Valiant 3 is a science, AI design, and general reasoning specialist built on Qwen 3.
- Finetuned on our newest science reasoning data generated with Deepseek R1 0528!
- AI to build AI: our high-difficulty AI reasoning data makes Shining Valiant 3 your friend for building with current AI tech and discovering new innovations and improvements!
- Improved general and creative reasoning to supplement problem-solving and general chat performance.
- Small model sizes allow running on local desktop and mobile, plus super-fast server inference!
Prompting Guide
Shining Valiant 3 uses the Qwen 3 prompt format.
Shining Valiant 3 is a reasoning finetune; we recommend enable_thinking=True for all chats.
Example inference script to get started:
from transformers import AutoModelForCausalLM, AutoTokenizer
model_name = "ValiantLabs/Qwen3-1.7B-ShiningValiant3"
# load the tokenizer and the model
tokenizer = AutoTokenizer.from_pretrained(model_name)
model = AutoModelForCausalLM.from_pretrained(
model_name,
torch_dtype="auto",
device_map="auto"
)
# prepare the model input
prompt = "Propose a novel cognitive architecture where the primary memory component is a Graph Neural Network (GNN). How would this GNN represent working, declarative, and procedural memory? How would the \"cognitive cycle\" be implemented as operations on this graph?"
messages = [
{"role": "user", "content": prompt}
]
text = tokenizer.apply_chat_template(
messages,
tokenize=False,
add_generation_prompt=True,
enable_thinking=True # Switches between thinking and non-thinking modes. Default is True.
)
model_inputs = tokenizer([text], return_tensors="pt").to(model.device)
# conduct text completion
generated_ids = model.generate(
**model_inputs,
max_new_tokens=32768
)
output_ids = generated_ids[0][len(model_inputs.input_ids[0]):].tolist()
# parsing thinking content
try:
# rindex finding 151668 (</think>)
index = len(output_ids) - output_ids[::-1].index(151668)
except ValueError:
index = 0
thinking_content = tokenizer.decode(output_ids[:index], skip_special_tokens=True).strip("\n")
content = tokenizer.decode(output_ids[index:], skip_special_tokens=True).strip("\n")
print("thinking content:", thinking_content)
print("content:", content)
Shining Valiant 3 is created by Valiant Labs.
Check out our HuggingFace page to see all of our models!
We care about open source. For everyone to use.
Model #2 - Lucy 128k (1.7B)
Lucy: Edgerunning Agentic Web Search on Mobile with a 1.7B model.
Authors: Alan Dao, Bach Vu Dinh, Alex Nguyen, Norapat Buppodom
Overview
Lucy is a compact but capable 1.7B model focused on agentic web search and lightweight browsing. Built on Qwen3-1.7B, Lucy inherits deep research capabilities from larger models while being optimized to run efficiently on mobile devices, even with CPU-only configurations.
We achieved this through machine-generated task vectors that optimize thinking processes, smooth reward functions across multiple categories, and pure reinforcement learning without any supervised fine-tuning.
What Lucy Excels At
- 🔍 Strong Agentic Search: Powered by MCP-enabled tools (e.g., Serper with Google Search)
- 🌐 Basic Browsing Capabilities: Through Crawl4AI (MCP server to be released), Serper,...
- 📱 Mobile-Optimized: Lightweight enough to run on CPU or mobile devices with decent speed
- 🎯 Focused Reasoning: Machine-generated task vectors optimize thinking processes for search tasks
Evaluation
Following the same MCP benchmark methodology used for Jan-Nano and Jan-Nano-128k, Lucy demonstrates impressive performance despite being only a 1.7B model, achieving higher accuracy than DeepSeek-v3 on SimpleQA.
🖥️ How to Run Locally
Lucy can be deployed using various methods including vLLM, llama.cpp, or through local applications like Jan, LMStudio, and other compatible inference engines. The model supports integration with search APIs and web browsing tools through the MCP.
Deployment
Deploy using VLLM:
vllm serve Menlo/Lucy-128k \
--host 0.0.0.0 \
--port 1234 \
--enable-auto-tool-choice \
--tool-call-parser hermes \
--rope-scaling '{"rope_type":"yarn","factor":3.2,"original_max_position_embeddings":40960}' --max-model-len 131072
Or llama-server from llama.cpp:
llama-server ... --rope-scaling yarn --rope-scale 3.2 --yarn-orig-ctx 40960
Recommended Sampling Parameters
Temperature: 0.7
Top-p: 0.9
Top-k: 20
Min-p: 0.0
🤝 Community & Support
- Discussions: HuggingFace Community
📄 Citation
Paper (coming soon): Lucy: edgerunning agentic web search on mobile with machine generated task vectors.
What is Brainstorm?
Brainstorm 20x
The BRAINSTORM process was developed by David_AU.
Some of the core principals behind this process are discussed in this scientific paper : Progressive LLaMA with Block Expansion .
However I went in a completely different direction from what was outlined in this paper.
What is "Brainstorm" ?
The reasoning center of an LLM is taken apart, reassembled, and expanded.
In this case for this model: 20 times
Then these centers are individually calibrated. These "centers" also interact with each other. This introduces subtle changes into the reasoning process. The calibrations further adjust - dial up or down - these "changes" further. The number of centers (5x,10x etc) allow more "tuning points" to further customize how the model reasons so to speak.
The core aim of this process is to increase the model's detail, concept and connection to the "world", general concept connections, prose quality and prose length without affecting instruction following.
This will also enhance any creative use case(s) of any kind, including "brainstorming", creative art form(s) and like case uses.
Here are some of the enhancements this process brings to the model's performance:
- Prose generation seems more focused on the moment to moment.
- Sometimes there will be "preamble" and/or foreshadowing present.
- Fewer or no "cliches"
- Better overall prose and/or more complex / nuanced prose.
- A greater sense of nuance on all levels.
- Coherence is stronger.
- Description is more detailed, and connected closer to the content.
- Simile and Metaphors are stronger and better connected to the prose, story, and character.
- Sense of "there" / in the moment is enhanced.
- Details are more vivid, and there are more of them.
- Prose generation length can be long to extreme.
- Emotional engagement is stronger.
- The model will take FEWER liberties vs a normal model: It will follow directives more closely but will "guess" less.
- The MORE instructions and/or details you provide the more strongly the model will respond.
- Depending on the model "voice" may be more "human" vs original model's "voice".
Other "lab" observations:
- This process does not, in my opinion, make the model 5x or 10x "smarter" - if only that was true!
- However, a change in "IQ" was not an issue / a priority, and was not tested or calibrated for so to speak.
- From lab testing it seems to ponder, and consider more carefully roughly speaking.
- You could say this process sharpens the model's focus on it's task(s) at a deeper level.
The process to modify the model occurs at the root level - source files level. The model can quanted as a GGUF, EXL2, AWQ etc etc.
For more information / other Qwen/Mistral Coders / additional settings see:
[ https://huggingface.co/DavidAU/Qwen2.5-MOE-2x-4x-6x-8x__7B__Power-CODER__19B-30B-42B-53B-gguf ]
Help, Adjustments, Samplers, Parameters and More
CHANGE THE NUMBER OF ACTIVE EXPERTS:
See this document:
https://huggingface.co/DavidAU/How-To-Set-and-Manage-MOE-Mix-of-Experts-Model-Activation-of-Experts
Settings: CHAT / ROLEPLAY and/or SMOOTHER operation of this model:
In "KoboldCpp" or "oobabooga/text-generation-webui" or "Silly Tavern" ;
Set the "Smoothing_factor" to 1.5
: in KoboldCpp -> Settings->Samplers->Advanced-> "Smooth_F"
: in text-generation-webui -> parameters -> lower right.
: In Silly Tavern this is called: "Smoothing"
NOTE: For "text-generation-webui"
-> if using GGUFs you need to use "llama_HF" (which involves downloading some config files from the SOURCE version of this model)
Source versions (and config files) of my models are here:
OTHER OPTIONS:
Increase rep pen to 1.1 to 1.15 (you don't need to do this if you use "smoothing_factor")
If the interface/program you are using to run AI MODELS supports "Quadratic Sampling" ("smoothing") just make the adjustment as noted.
Highest Quality Settings / Optimal Operation Guide / Parameters and Samplers
This a "Class 1" model:
For all settings used for this model (including specifics for its "class"), including example generation(s) and for advanced settings guide (which many times addresses any model issue(s)), including methods to improve model performance for all use case(s) as well as chat, roleplay and other use case(s) please see:
You can see all parameters used for generation, in addition to advanced parameters and samplers to get the most out of this model here: