Instructions to use dusersad12/QuasarLM-ReleaseRepo with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use dusersad12/QuasarLM-ReleaseRepo with Transformers:

# Use a pipeline as a high-level helper
from transformers import pipeline

pipe = pipeline("text-generation", model="dusersad12/QuasarLM-ReleaseRepo")

# Load model directly
from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("dusersad12/QuasarLM-ReleaseRepo")
model = AutoModelForCausalLM.from_pretrained("dusersad12/QuasarLM-ReleaseRepo", device_map="auto")

Notebooks
Google Colab
Kaggle
Local Apps Settings

vLLM

How to use dusersad12/QuasarLM-ReleaseRepo with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "dusersad12/QuasarLM-ReleaseRepo"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "dusersad12/QuasarLM-ReleaseRepo",
		"prompt": "Once upon a time,",
		"max_tokens": 512,
		"temperature": 0.5
	}'

Use Docker

docker model run hf.co/dusersad12/QuasarLM-ReleaseRepo

SGLang

How to use dusersad12/QuasarLM-ReleaseRepo with SGLang:

Install from pip and serve model

# Install SGLang from pip:
pip install sglang
# Start the SGLang server:
python3 -m sglang.launch_server \
    --model-path "dusersad12/QuasarLM-ReleaseRepo" \
    --host 0.0.0.0 \
    --port 30000
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:30000/v1/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "dusersad12/QuasarLM-ReleaseRepo",
		"prompt": "Once upon a time,",
		"max_tokens": 512,
		"temperature": 0.5
	}'

Use Docker images

docker run --gpus all \
    --shm-size 32g \
    -p 30000:30000 \
    -v ~/.cache/huggingface:/root/.cache/huggingface \
    --env "HF_TOKEN=<secret>" \
    --ipc=host \
    lmsysorg/sglang:latest \
    python3 -m sglang.launch_server \
        --model-path "dusersad12/QuasarLM-ReleaseRepo" \
        --host 0.0.0.0 \
        --port 30000
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:30000/v1/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "dusersad12/QuasarLM-ReleaseRepo",
		"prompt": "Once upon a time,",
		"max_tokens": 512,
		"temperature": 0.5
	}'

Docker Model Runner
How to use dusersad12/QuasarLM-ReleaseRepo with Docker Model Runner:
```
docker model run hf.co/dusersad12/QuasarLM-ReleaseRepo
```

QuasarLM

1. Introduction

QuasarLM is the newest open release from Northwind AI Research. This cycle the model went through a substantial post-training overhaul: longer reasoning chains, a heavier curriculum of verifiable-reward problems, and a revised instruction-mix during supervised fine-tuning. The result is a compact model that holds its own on mathematics, coding and general logic evaluations, with an overall profile that now tracks noticeably stronger open models in the same size class.

Compared to the previous internal iteration, the biggest movement is on complex multi-step reasoning. On the AIME 2026 test, for example, accuracy has climbed from 64% for the previous iteration to 81.2% for this release. The gain lines up with how much more the model thinks before it answers: where the earlier version averaged 9K tokens per AIME question, the current release averages 19K.

Alongside the reasoning gains, this release also ships a lower hallucination rate and expanded tool-calling support.

2. Evaluation Results

Comprehensive Benchmark Results

	Benchmark	Nimbus-3B	Solaris-7B	Nimbus-3B-v2	QuasarLM
Core Reasoning Tasks	Math Reasoning	0.626	0.642	0.657	0.668
	Logical Reasoning	0.702	0.719	0.736	0.748
	Common Sense	0.615	0.631	0.645	0.655
Language Understanding	Reading Comprehension	0.591	0.608	0.623	0.635
	Question Answering	0.599	0.616	0.631	0.642
	Text Classification	0.695	0.712	0.729	0.742
	Sentiment Analysis	0.671	0.687	0.701	0.712
Generation Tasks	Code Generation	0.542	0.559	0.573	0.585
	Creative Writing	0.506	0.498	0.531	0.551
	Dialogue Generation	0.508	0.525	0.540	0.552
	Summarization	0.606	0.622	0.637	0.648
Specialized Capabilities	Translation	0.686	0.703	0.719	0.731
	Knowledge Retrieval	0.547	0.563	0.578	0.590
	Instruction Following	0.652	0.669	0.686	0.698
	Safety Evaluation	0.661	0.679	0.696	0.709

Overall Performance Summary

QuasarLM delivers consistent headroom across all evaluated benchmark categories, with the largest margins concentrated in the reasoning and generation clusters.

3. Chat Website & API Platform

We host a playground and serve an API so anyone can chat with QuasarLM. Check our official website for the details.

4. How to Run Locally

Please refer to our code repository for more information about running QuasarLM locally.

Compared to previous versions, the usage recommendations for QuasarLM have the following changes:

System prompt is supported.
It is not required to add special tokens at the beginning of the output to force the model into a specific thinking pattern.

The model architecture of QuasarLM-Small is identical to its base model, but it shares the same tokenizer configuration as the main QuasarLM. This model can be run in the same manner as its base model.

System Prompt

We recommend using the following system prompt with a specific date.

You are QuasarLM, a helpful AI assistant.
Today is {current date}.

For example,

You are QuasarLM, a helpful AI assistant.
Today is September 22, 2026, Tuesday.

Temperature

We recommend setting the temperature parameter $T_{model}$ to 0.6.

Prompts for File Uploading and Web Search

For file uploading, please follow the template to create prompts, where {file_name}, {file_content} and {question} are arguments.

file_template = \
"""[file name]: {file_name}
[file content begin]
{file_content}
[file content end]
{question}"""

For web search enhanced generation, we recommend the following prompt template where {search_results}, {cur_date}, and {question} are arguments.

search_answer_en_template = \
'''# The following contents are the search results related to the user's message:
{search_results}
In the search results I provide to you, each result is formatted as [webpage X begin]...[webpage X end], where X represents the numerical index of each article. Please cite the context at the end of the relevant sentence when appropriate. Use the citation format [citation:X] in the corresponding part of your answer. If a sentence is derived from multiple contexts, list all relevant citation numbers, such as [citation:3][citation:5]. Be sure not to cluster all citations at the end; instead, include them in the corresponding parts of the answer.
When responding, please keep the following points in mind:
- Today is {cur_date}.
- Not all content in the search results is closely related to the user's question. You need to evaluate and filter the search results based on the question.
- For listing-type questions (e.g., listing all flight information), try to limit the answer to 10 key points and inform the user that they can refer to the search sources for complete information. Prioritize providing the most complete and relevant items in the list. Avoid mentioning content not provided in the search results unless necessary.
- For creative tasks (e.g., writing an essay), ensure that references are cited within the body of the text, such as [citation:3][citation:5], rather than only at the end of the text. You need to interpret and summarize the user's requirements, choose an appropriate format, fully utilize the search results, extract key information, and generate an answer that is insightful, creative, and professional. Extend the length of your response as much as possible, addressing each point in detail and from multiple perspectives, ensuring the content is rich and thorough.
- If the response is lengthy, structure it well and summarize it in paragraphs. If a point-by-point format is needed, try to limit the answer to 5 points and merge related content.
- For objective Q&A, if the answer is very brief, you may add one or two related sentences to enrich the content.
- Choose an appropriate and visually appealing format for your response based on the user's requirements and the content of the answer, ensuring strong readability.
- Your answer should synthesize information from multiple relevant webpages and avoid repeatedly citing the same webpage.
- Unless the user requests otherwise, your response should be in the same language as the user's question.
# The user's message is:
{question}'''

5. License

This code repository is licensed under the Apache 2.0 License. The use of QuasarLM models is also subject to the Apache 2.0 License. The model series supports commercial use and distillation.

6. Contact

If you have any questions, please raise an issue on our GitHub repository or contact us at contact@quasarlm.ai.

Downloads last month: -