# AhoosAI > AhoosAI develops the Nimbus language models and provides them through an online studio, an offline studio for your own computer, and a public API compatible with OpenAI's Chat Completions format (base URL https://ahoos-ai.site/services/api/v1, model `nimbus-2-apex`). The documentation is published as Markdown in English and Persian. Each news article has a static page containing both languages. https://ahoos-ai.site/llms-full.txt holds the whole documentation in one file. ## Documentation - [AhoosAI documentation](https://ahoos-ai.site/docs/index.en.txt): AhoosAI develops the Nimbus language models and provides them through three products: an online studio, an offline studio for your own computer, and a public API compatible with OpenAI's Chat Completions format. This documentation describes what you need to use them. - [API reference](https://ahoos-ai.site/docs/api.en.txt): The AhoosAI API serves the Nimbus 2 Apex model in the format of OpenAI's Chat Completions API. Existing OpenAI client libraries work with it once the base URL and the key are changed. - [API rules and limits](https://ahoos-ai.site/docs/rules.en.txt): These rules apply to the AhoosAI API and to the online studio. They complement the Terms of use and the Privacy notice. By using the API you accept them. - [Models](https://ahoos-ai.site/docs/models.en.txt): Nimbus 2 Apex is a LoRA adapter over Qwen3-8B, trained to follow instructions closely. It reasons before it answers, and two settings control how it does so. - [Studio](https://ahoos-ai.site/docs/studio.en.txt): AhoosAI Studio comes in two forms: an online studio in the browser, and an offline studio that runs on your own computer. ## Documentation in Persian - [مستندات اهوس‌ای‌آی](https://ahoos-ai.site/docs/index.fa.txt): اهوس‌ای‌آی مدل‌های زبانی نیمبوس را توسعه می‌دهد و آن‌ها را از راه سه محصول ارائه می‌کند: استودیوی آنلاین، استودیوی آفلاین برای اجرا روی رایانهٔ شخصی، و یک API عمومی سازگار با قالب Chat Completions شرکت OpenAI. این مستندات آنچه را برای استفاده از این محصولات لازم است شرح می‌دهد. - [مرجع API](https://ahoos-ai.site/docs/api.fa.txt): API اهوس‌ای‌آی مدل نیمبوس ۲ ایپکس را در قالب Chat Completions شرکت OpenAI ارائه می‌کند. کتابخانه‌های موجود OpenAI پس از تغییر نشانی پایه و کلید، بدون تغییر دیگری با آن کار می‌کنند. - [قوانین و محدودیت‌های API](https://ahoos-ai.site/docs/rules.fa.txt): این قوانین بر API اهوس‌ای‌آی و استودیوی آنلاین حاکم‌اند و مکمل شرایط استفاده و سیاست حریم خصوصی هستند. استفاده از API به معنای پذیرش این قوانین است. - [مدل‌ها](https://ahoos-ai.site/docs/models.fa.txt): نیمبوس ۲ ایپکس یک آداپتر LoRA روی مدل Qwen3-8B است که برای پیروی دقیق از دستورها آموزش دیده است. این مدل پیش از پاسخ استدلال می‌کند و دو تنظیم، چگونگی آن را تعیین می‌کنند. - [استودیو](https://ahoos-ai.site/docs/studio.fa.txt): استودیوی اهوس‌ای‌آی در دو شکل ارائه می‌شود: استودیوی آنلاین در مرورگر، و استودیوی آفلاین که روی رایانهٔ خود شما اجرا می‌شود. ## News - [AhoosAI Studio 3.1, PML-sRUN out under MIT, and the Terminal CLI](https://ahoos-ai.site/news/studio-3-1.html): Three releases in one day: the offline studio twice as fast on a small graphics card and working with any model you have a key for; PML-sRUN, the method behind it, open on GitHub with a control panel; and the Terminal CLI, with the online studio's abilities in a terminal and an agent of its own. - [PML-sRUN: a model bigger than the machine, run in pieces](https://ahoos-ai.site/news/pml-srun.html): PML-sRUN made by AhoosAI: a model on hardware it does not fit, each piece of it where it runs fastest. Now open under the MIT license, with a control panel -- and on a 4 GB card, twice as fast as llama.cpp on its own. - [Nimbus 2 Apex is online](https://ahoos-ai.site/news/nimbus-2-apex-online.html): Nimbus 2 Apex now answers through the AhoosAI API and the studio, on an NVIDIA T4 graphics card, run with PML-sRUN. The offline studio gets it soon. - [AhoosAI Studio 3.0: a model that finishes the job](https://ahoos-ai.site/news/studio-3-0.html): Ask for a project and the model plans it, writes every file in full, checks what it wrote and fixes it before answering. One click undoes a whole answer, servers keep running in the background, and the model runs on your graphics card. - [AhoosAI Studio 2.5: the model does the work](https://ahoos-ai.site/news/studio-2-5.html): Connect a folder and the model writes the files and runs the commands itself, asking you before every change. Also new: internet access if you want it, a battle between the base model and its adapter, and any model from Hugging Face. - [AhoosAI Studio runs Nimbus 2 Apex on your own computer](https://ahoos-ai.site/news/offline-studio.html): One download, then nothing leaves the machine. The two dials the model was trained under, memory between messages, a folder it may work in, and a gate in front of everything that touches it. - [Nimbus 2 Apex: one gain, and everything that did not move](https://ahoos-ai.site/news/nimbus-2-apex.html): Instruction-following up 8.7 points against the model it was trained on, and no loss anywhere that survives a significance test. The thinking ladder did not work, and the reason is in the data. - [Nimbus 1.1 Prime-EE, measured and published](https://ahoos-ai.site/news/nimbus-1-1-prime.html): The first model of AhoosAI’s own. The scores, what moved against the same weights with the adapter off, and the one area that lost ground — with the weights on Hugging Face. - [AhoosAI Studio enters beta](https://ahoos-ai.site/news/studio-beta.html): A chat client for the Nimbus family: sign in with Google or GitHub, keep your history, watch the manager hand work to a specialist, and take away the files it wrote. - [Nimbus answering real requests](https://ahoos-ai.site/news/nimbus-live.html): The manager-and-specialists loop running outside a test harness: planning, delegation and gear routing all working end to end. - [The image model draws](https://ahoos-ai.site/news/image-model.html): Until now the image specialist produced written specifications. It now returns pictures, and the manager routes to it whenever a request calls for one. - [Two hundred evaluation tasks](https://ahoos-ai.site/news/eval-200.html): Four suites asking one question: does the layer earn its cost against the plain model underneath? A result at or below zero is a finding, and we publish those too. ## Optional - [Models](https://ahoos-ai.site/models/): the model catalogue. - [AhoosAI Studio, on your own computer](https://ahoos-ai.site/news/int-pages/offline-studio.html): the offline studio and its download. - [Terms of use](https://ahoos-ai.site/services/studio/terms.html) and [Privacy notice](https://ahoos-ai.site/services/studio/privacy.html).