This repo contains the full precision source code, in "safe tensors" format to generate GGUFs, GPTQ, EXL2, AWQ, HQQ and other formats. The source code can also be used directly.
This model is for all use cases, but excels in CODING and programming in all major programming languages and many minor ones too.
This model is based on Qwen3-30B-A3B-Instruct-2507 (MOE, 128 experts, 8 activated), with Brainstorm 20X (by DavidAU) - details at bottom of this page.
The Brainstorm adapter will improve general performance and "out of the box" thinking.
This creates a model of 42B parameters, 67 layers and 807 tensors.
This version has the NATIVE context of 256k.
This version (vs Brainstorm 40x, 53B version) can be used for both coding and creative use cases.
For coding, programming set expert to:
Recommended settings - general:
Creative Use Cases:
This is the refined version -V1.4- from this project (see this repo for all settings, details, system prompts, example generations etc etc):
https://huggingface.co/DavidAU/Qwen3-55B-A3B-TOTAL-RECALL-Deep-40X-GGUF/
This version 2 is slightly smaller, with further refinements to the Brainstorm adapter and uses the new "Qwen3-30B-A3B-Instruct-2507".
Review and Specialized Settings for this model (V 1.4):
https://www.linkedin.com/posts/gchesler_haskell-postgres-agentic-activity-7347103276141596672-_zbo/
You may also want to see (root model of Total Recall series - Version 1):
https://huggingface.co/Qwen/Qwen3-30B-A3B
AND Version 2 root model:
https://huggingface.co/Qwen/Qwen3-30B-A3B-Thinking-2507
For additional settings, tool use, and other model settings.
Summary of root model below, followed by FULL HELP SECTION, then info on Brainstorm 40x.
QUANTS
Special Thanks to Team Mradermacher, and Nightmedia for the quants:
GGUF:
GGUF-IMATRIX:
MLX:
Over the past three months, we have continued to scale the thinking capability of Qwen3-30B-A3B, improving both the quality and depth of reasoning. We are pleased to introduce Qwen3-30B-A3B-Thinking-2507, featuring the following key enhancements:
NOTE: This version has an increased thinking length. We strongly recommend its use in highly complex reasoning tasks.
Qwen3-30B-A3B-Thinking-2507 has the following features:
NOTE: This model supports only thinking mode. Meanwhile, specifying enable_thinking=True is no longer required.
Additionally, to enforce model thinking, the default chat template automatically includes <think>. Therefore, it is normal for the model's output to contain only </think> without an explicit opening <think> tag.
For more details, including benchmark evaluation, hardware requirements, and inference performance, please refer to our blog, GitHub, and Documentation.
| Gemini2.5-Flash-Thinking | Qwen3-235B-A22B Thinking | Qwen3-30B-A3B Thinking | Qwen3-30B-A3B-Thinking-2507 | |
|---|---|---|---|---|
| Knowledge | ||||
| MMLU-Pro | 81.9 | 82.8 | 78.5 | 80.9 |
| MMLU-Redux | 92.1 | 92.7 | 89.5 | 91.4 |
| GPQA | 82.8 | 71.1 | 65.8 | 73.4 |
| SuperGPQA | 57.8 | 60.7 | 51.8 | 56.8 |
| Reasoning | ||||
| AIME25 | 72.0 | 81.5 | 70.9 | 85.0 |
| HMMT25 | 64.2 | 62.5 | 49.8 | 71.4 |
| LiveBench 20241125 | 74.3 | 77.1 | 74.3 | 76.8 |
| Coding | ||||
| LiveCodeBench v6 (25.02-25.05) | 61.2 | 55.7 | 57.4 | 66.0 |
| CFEval | 1995 | 2056 | 1940 | 2044 |
| OJBench | 23.5 | 25.6 | 20.7 | 25.1 |
| Alignment | ||||
| IFEval | 89.8 | 83.4 | 86.5 | 88.9 |
| Arena-Hard v2$ | 56.7 | 61.5 | 36.3 | 56.0 |
| Creative Writing v3 | 85.0 | 84.6 | 79.1 | 84.4 |
| WritingBench | 83.9 | 80.3 | 77.0 | 85.0 |
| Agent | ||||
| BFCL-v3 | 68.6 | 70.8 | 69.1 | 72.4 |
| TAU1-Retail | 65.2 | 54.8 | 61.7 | 67.8 |
| TAU1-Airline | 54.0 | 26.0 | 32.0 | 48.0 |
| TAU2-Retail | 66.7 | 40.4 | 34.2 | 58.8 |
| TAU2-Airline | 52.0 | 30.0 | 36.0 | 58.0 |
| TAU2-Telecom | 31.6 | 21.9 | 22.8 | 26.3 |
| Multilingualism | ||||
| MultiIF | 74.4 | 71.9 | 72.2 | 76.4 |
| MMLU-ProX | 80.2 | 80.0 | 73.1 | 76.4 |
| INCLUDE | 83.9 | 78.7 | 71.9 | 74.4 |
| PolyMATH | 49.8 | 54.7 | 46.1 | 52.6 |
$ For reproducibility, we report the win rates evaluated by GPT-4.1.
& For highly challenging tasks (including PolyMATH and all reasoning and coding tasks), we use an output length of 81,920 tokens. For all other tasks, we set the output length to 32,768.
The code of Qwen3-MoE has been in the latest Hugging Face transformers and we advise you to use the latest version of transformers.
With transformers<4.51.0, you will encounter the following error:
KeyError: 'qwen3_moe'
The following contains a code snippet illustrating how to use the model generate content based on given inputs.
from transformers import AutoModelForCausalLM, AutoTokenizer
model_name = "Qwen/Qwen3-30B-A3B-Thinking-2507"
# 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 = "Give me a short introduction to large language model."
messages = [
{"role": "user", "content": prompt}
]
text = tokenizer.apply_chat_template(
messages,
tokenize=False,
add_generation_prompt=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) # no opening <think> tag
print("content:", content)
For deployment, you can use sglang>=0.4.6.post1 or vllm>=0.8.5 or to create an OpenAI-compatible API endpoint:
python -m sglang.launch_server --model-path Qwen/Qwen3-30B-A3B-Thinking-2507 --context-length 262144 --reasoning-parser deepseek-r1
vllm serve Qwen/Qwen3-30B-A3B-Thinking-2507 --max-model-len 262144 --enable-reasoning --reasoning-parser deepseek_r1
Note: If you encounter out-of-memory (OOM) issues, you may consider reducing the context length to a smaller value. However, since the model may require longer token sequences for reasoning, we strongly recommend using a context length greater than 131,072 when possible.
For local use, applications such as Ollama, LMStudio, MLX-LM, llama.cpp, and KTransformers have also supported Qwen3.
Qwen3 excels in tool calling capabilities. We recommend using Qwen-Agent to make the best use of agentic ability of Qwen3. Qwen-Agent encapsulates tool-calling templates and tool-calling parsers internally, greatly reducing coding complexity.
To define the available tools, you can use the MCP configuration file, use the integrated tool of Qwen-Agent, or integrate other tools by yourself.
from qwen_agent.agents import Assistant
# Define LLM
# Using Alibaba Cloud Model Studio
llm_cfg = {
'model': 'qwen3-30b-a3b-thinking-2507',
'model_type': 'qwen_dashscope',
}
# Using OpenAI-compatible API endpoint. It is recommended to disable the reasoning and the tool call parsing
# functionality of the deployment frameworks and let Qwen-Agent automate the related operations. For example,
# `VLLM_USE_MODELSCOPE=true vllm serve Qwen/Qwen3-30B-A3B-Thinking-2507 --served-model-name Qwen3-30B-A3B-Thinking-2507 --tensor-parallel-size 8 --max-model-len 262144`.
#
# llm_cfg = {
# 'model': 'Qwen3-30B-A3B-Thinking-2507',
#
# # Use a custom endpoint compatible with OpenAI API:
# 'model_server': 'http://localhost:8000/v1', # api_base without reasoning and tool call parsing
# 'api_key': 'EMPTY',
# 'generate_cfg': {
# 'thought_in_content': True,
# },
# }
# Define Tools
tools = [
{'mcpServers': { # You can specify the MCP configuration file
'time': {
'command': 'uvx',
'args': ['mcp-server-time', '--local-timezone=Asia/Shanghai']
},
"fetch": {
"command": "uvx",
"args": ["mcp-server-fetch"]
}
}
},
'code_interpreter', # Built-in tools
]
# Define Agent
bot = Assistant(llm=llm_cfg, function_list=tools)
# Streaming generation
messages = [{'role': 'user', 'content': 'https://qwenlm.github.io/blog/ Introduce the latest developments of Qwen'}]
for responses in bot.run(messages=messages):
pass
print(responses)
To achieve optimal performance, we recommend the following settings:
Sampling Parameters:
Temperature=0.6, TopP=0.95, TopK=20, and MinP=0.presence_penalty parameter between 0 and 2 to reduce endless repetitions. However, using a higher value may occasionally result in language mixing and a slight decrease in model performance.Adequate Output Length: We recommend using an output length of 32,768 tokens for most queries. For benchmarking on highly complex problems, such as those found in math and programming competitions, we suggest setting the max output length to 81,920 tokens. This provides the model with sufficient space to generate detailed and comprehensive responses, thereby enhancing its overall performance.
Standardize Output Format: We recommend using prompts to standardize model outputs when benchmarking.
answer field with only the choice letter, e.g., "answer": "C"."No Thinking Content in History: In multi-turn conversations, the historical model output should only include the final output part and does not need to include the thinking content. It is implemented in the provided chat template in Jinja2. However, for frameworks that do not directly use the Jinja2 chat template, it is up to the developers to ensure that the best practice is followed.
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:
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:
Other "lab" observations:
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.