dusersad12/NimbusModel-ReleaseRepo

NimbusModel

---

1. Introduction

NimbusModel just went through a major version bump. In this release, NimbusModel gains much deeper reasoning and inference abilities thanks to a larger compute budget and new algorithmic optimizations introduced during post-training. The model delivers strong results across a wide range of benchmark evaluations, covering mathematics, programming, and general logic, and its overall performance is now closing in on several leading models.

Compared with the previous release, the upgraded model shows clear gains on complex reasoning tasks. On the AIME 2025 test set, accuracy moved from 68% in the previous release to 89.3% in the current one. The improvement comes from a longer thinking process: where the older model spent about 10K tokens per question on AIME, the new release averages 21K tokens per question.

Alongside the stronger reasoning, this release also cuts the hallucination rate and improves function calling support.

2. Evaluation Results

Comprehensive Benchmark Results

	Benchmark	ModelA	ModelB	ModelA-v2	NimbusModel
Core Reasoning Tasks	Math Reasoning	0.492	0.517	0.503	0.858
	Logical Reasoning	0.758	0.771	0.782	0.881
	Common Sense	0.694	0.681	0.704	0.706
Language Understanding	Reading Comprehension	0.648	0.662	0.667	0.656
	Question Answering	0.561	0.578	0.580	0.824
	Text Classification	0.788	0.796	0.805	0.754
	Sentiment Analysis	0.752	0.756	0.765	0.739
Generation Tasks	Code Generation	0.597	0.613	0.622	0.832
	Creative Writing	0.570	0.561	0.583	0.694
	Dialogue Generation	0.602	0.616	0.620	0.677
	Summarization	0.721	0.731	0.736	0.754
Specialized Capabilities	Translation	0.756	0.773	0.775	0.766
	Knowledge Retrieval	0.628	0.645	0.647	0.671
	Instruction Following	0.710	0.726	0.728	0.855
	Safety Evaluation	0.695	0.678	0.702	0.841

Overall Performance Summary

The NimbusModel shows strong performance across all evaluated benchmark categories, with particularly notable results in reasoning and generation tasks.

3. Chat Website & API Platform

You can try NimbusModel through our chat interface and API platform; see the official website for details.

4. How to Run Locally

For instructions on running NimbusModel locally, please refer to our code repository.

Compared with earlier releases, the usage recommendations for NimbusModel changed as follows:

1. A system prompt is supported.
2. You no longer need to prepend special tokens to the output to force a specific thinking pattern.

The model architecture of NimbusModel-Small is identical to its base model, but it shares the same tokenizer configuration as the main NimbusModel. 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 NimbusModel, a helpful AI assistant.
Today is {current date}.

For example,

You are NimbusModel, a helpful AI assistant.
Today is September 27, 2026, Sunday.

Temperature

We suggest setting the temperature parameter $T_{model}$ to 0.55.

Prompts for File Uploading and Web Search

When uploading files, build the prompt with the following template, 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 user's requirements.
- 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 included 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 items.
- 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 answer 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 NimbusModel 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@nimbusmodel.ai.