dusersad12/NexusModel-EvalRepo

NexusModel

---

1. Introduction

The NexusModel has just gone through a major mid-cycle refresh. With the latest update, NexusModel substantially deepens its reasoning and inference strength by drawing on additional compute and a new algorithmic optimization stack during post-training. The model posts competitive numbers across a wide range of benchmark evaluations, spanning mathematics, coding, and general logic. Its overall capability profile is now closing in on the strongest models in its class.

Compared with the previous iteration, the refreshed model handles complex multi-step reasoning noticeably better. For instance, on the AIME 2025 evaluation set, accuracy rose from 68% in the prior release to 85% in this one. The gain comes from longer thinking depth during reasoning: in the AIME test set, the old model averaged around 11K tokens per question, while the new build averages 21K tokens per question.

Alongside the stronger reasoning, this release also brings a lower hallucination rate and better function-calling support.

2. Evaluation Results

Comprehensive Benchmark Results

	Benchmark	ModelA	ModelB	ModelA-v2	NexusModel
Core Reasoning Tasks	Math Reasoning	0.498	0.521	0.512	0.610
	Logical Reasoning	0.772	0.785	0.796	0.830
	Common Sense	0.701	0.690	0.713	0.750
Language Understanding	Reading Comprehension	0.658	0.672	0.678	0.710
	Question Answering	0.571	0.588	0.590	0.620
	Text Classification	0.789	0.797	0.806	0.840
	Sentiment Analysis	0.763	0.769	0.777	0.800
Generation Tasks	Code Generation	0.603	0.619	0.628	0.656
	Creative Writing	0.576	0.569	0.590	0.617
	Dialogue Generation	0.609	0.623	0.627	0.660
	Summarization	0.731	0.742	0.747	0.780
Specialized Capabilities	Translation	0.768	0.783	0.787	0.820
	Knowledge Retrieval	0.638	0.655	0.657	0.690
	Instruction Following	0.720	0.735	0.738	0.770
	Safety Evaluation	0.705	0.690	0.713	0.750
Extended Capabilities	Tool Use	0.582	0.601	0.615	0.700
	Multimodal Reasoning	0.595	0.612	0.625	0.690

Overall Performance Summary

The NexusModel shows solid results in every evaluated benchmark category, with the clearest headroom gained in reasoning and generation tasks.

3. Chat Website & API Platform

We run a chat interface and an API for talking to NexusModel. Check our official website for details.

4. How to Run Locally

Head over to our code repository for instructions on running NexusModel locally.

Relative to the last release, the usage guidance for NexusModel changes as follows:

1. A system prompt is supported.
2. Special tokens are no longer needed at the start of the output to force a particular thinking mode.

The architecture of NexusModel-Small matches its base model exactly, and it shares the tokenizer configuration of the main NexusModel. Run it the same way you would run its base model.

System Prompt

We suggest the following system prompt with a concrete date.

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

For example,

You are NexusModel, a helpful AI assistant.
Today is August 15, 2026, Friday.

Temperature

We suggest running with the temperature parameter $T_{model}$ set to 0.6.

Prompts for File Uploading and Web Search

For file uploads, build prompts following the template below, 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-assisted generation, use 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 it 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. Use of the NexusModel models is likewise governed by the Apache 2.0 License. The model family supports commercial use and distillation.

6. Contact

For questions, open an issue on our GitHub repository or reach us at contact@nexusmodel.ai.