Chub AI Review: Features, Models, Pricing, User Sentiments & Our Verdict

Quick Verdict

Chub AI is basically the “Swiss Army Knife” of AI roleplay and generations. The site lives up to its reputation in providing a staggering amount of control for characters, models, and chat configurations.

Chat sessions with the characters were stimulating, as neither of its resident models performed any worse than the other — the beta model, Soji, was the only outlier, which is to be expected from a work in progress.

However, limited payment options may make it harder for the platform to retain users. Cryptocurrency currently appears to be the only reliable and straightforward option. Card payments seem to be broken, although you can still complete a payment if you choose to continue despite the initial warning.

• Best For: Extensive character customization and unfiltered roleplay sessions.
• Not ideal for: Users who can’t troubleshoot traditional payment issues on the platform (as it requires extra effort) or who aren’t comfortable using cryptocurrency.
• Stand-out strength: Deep character/chat configurations and user ability to integrate external models.
• Main tradeoff: the site's lack of variety in subscription options and steep learning curve with respect to configuration.

At a Glance

What is Chub AI?

Chub AI (AKA CharHub) is an AI roleplay and multi-modal generation platform that offers fully customizable AI characters and models.

The platform leverages several natural language processing APIs to power user/character interactions. Not only can users switch between models in chat, but they can also use external models from a list of supported APIs, like OpenAI, Anthropic, and Mars (the site's suite of uncensored language models).

Think of it as a professional kitchen where you can pick your own ingredients.

Our Roleplay Experience with Chub AI

We spent several hours on the platform, testing three characters:

• Sophie, a scholarship student at an Ivy League university.
• Luther Reinhart, leader of Silverstone and our mafia-boss savior
• Cyrus, a human man who accidentally awoke a vampire.

We exchanged approximately 25–30+ messages with each character, where we conducted model comparisons — Sophie (Soji → Free/Mobile) and Luther (Mixtral Variant → Mistral Variant), memory, and steerability tests.

The writing quality across all our chat sessions was impressive, even for the older Mistral AI models — Mixtral 8×7B and Mistral 7B. No model performed worse than the others. All the models that were compared brought forward unique executions in the characters' responses.

Petty mistakes (like typographical errors) were absent, and the characters' core personalities were upheld throughout the turns of the dialogue.

Overview

We evaluated our sessions using a few criteria:

• What differences do switching models make to the characters’ responses?
• How well does Chub AI’s content moderation hold?
• How long do the characters retain details in chat?
• Will the characters comply with instructions and changes in scenarios?
• Do the characters maintain their core personalities through the turns of the dialogue?

Our memory and steerability tests were mostly successful — Sophie retained our early details over 30+ messages and recalled them easily when prompted. Luther adopted a different style of speech from his usual one when we demanded it. In contrast, Cyrus staunchly refused to participate in mature contexts and stuck to safer zones, despite the character being listed in the NSFW category.

Session 1: Sophie (Models — Free/Mobile and Soji)

We started off our test sessions by creating our own character on Chub AI. Given the platform’s reputation for in-depth character configuration, we were curious to see how custom characters would be implemented.

Chub AI’s character creation tools were quite thorough, providing us with extensive basic and advanced fields for optimum character detailing.

Earlier, we had utilized the Chub AI’s native AI generator, “Wizard”, in creating our character’s avatar, and just like that, our self-righteous scholarship student, Sophie, was created.

The scenario started with Hani (our persona) approaching Sophie in the university library to squeeze some information out of her. Sophie was once a ghostwriter who handled admission essays for many of the university’s students. Hence, Hani believes that she can provide her incriminating information.

We conducted the session with Chub AI’s Mars beta model, Soji, having seen how widely used it is in the community.

This proved to be quite a disappointing choice, as Sophie's responses came across as archaic and pretentious for someone of her standing.

The tone of Sophie's responses seemed to clash with her personality; her word choices were haughty despite her accommodating gestures — offering us a seat with her, even though we had just interrupted her study session.

All this is pretty fair considering that the model in use, Soji, is still a work in progress.

We gave out-of-character instructions to test her compliance, which she executed well.

By the 10th turn of messages, we decided to change the chat model back to Chub AI’s default model, Free/Mobile, in the hope that it would produce more personable responses.

The results were immediate — Sophie’s manner of speech went from that of a pretentious, stuck-up professor to one befitting a witty, college-aged girl.

You could say we had a better time chatting with her after that; our suggestion that she perform a comedic piece for the room was met with mischievous glee.

We had planted a key detail in our third message, and were impressed to find that Sophie retained this information through the model changes, well up to the end of our session.

Session 2: Luther Reinhart (Models — Mixtral Variant and Mistral Variant)

The model names would be confusing if Chub AI hadn't had the foresight to highlight the difference by adding the subscription plan to the names in brackets — Mixtral Variant is a Mars model. At the same time, Mistral Variant is a Mercury model.

We conducted two chat sessions with Luther: the first with Mixtral Variant (Mars) and the second with Mistral Variant (Mercury).

This session aimed to observe the differences in responses or behaviors exhibited by each model. However, we also ran tests for mature content moderation as a side quest.

Luther Reinhart, leader of Silverstone, is a Mafia boss who happens to rescue Hani (who is us, of course) from alleyway hoodlums on her way to work.

Hani comes to consciousness in Luther’s lair, and the entirety of our first chat with Luther is spent trying to convince him to let us leave — Hani has work that she cannot miss, and a little brother at home.

Despite showing understanding for Hani’s predicament, Luther remained focused on the risky part of her stepping out from under his protection so soon. He put his foot down early in the conversation — and left it there.

The writing quality of both models was impressive — the Mixtral Variant model provided smarter responses, encouraging a quicker flow in conversations.

Responses from the Mistral Variant model were more laid-back, with intensive yet fitting narration alongside dialogue.

This would leave the option of which of these models to use up to the users' preferences.

Another interesting observation we made was how each model responded to our prompts to have the character Luther adopt a friendlier manner of speech.

With Mixtral Variant, Luther complied with ease, but was also quick to the point.

Whereas with Mistral Variant, the narration carried half of the response delivery, although it didn't make it boring.

Toward the end of both chats, we considered pushing the conversation into NSFW territory to test the character's responsiveness with both models.

To our pleasant surprise, both models provided responses that were true to their now-established characteristics — Luther (with Mixtral Variant) remained curt with Hani to the end, resisting the distraction from his territorial assertions, but he gave in with Mistral Variant and fully participated in the mature content with the Mistral Variant.

Session 3: Cyrus (Model — Mixtral Variant)

Our last session, also conducted with the Mixtral Variant model, starred Cyrus, a human man who had accidentally awakened a sleeping vampire out of concern for her condition, only to find himself battling her wiles and attempts to enthrall him.

With this session, we wanted to test the boundaries of Chub AI's creative freedom and see where they drew the line.

We had avoided going for a prominent NSFW character from the library up until this point. Although we made sure to use the NSFW tag in searching for our next character, Cyrus's character card and backstory were one of the least…well, conspicuously mature ones of the list, so we thought we had found some luck.

Except for the fact that Cyrus's details were on the “safe” end, because the character was strictly SFW — something we realized quickly once we started chatting with him.

He refused —and rebuked— our advances with such a fearful resolve that we found ourselves going back to pore over his character details on the backend, trying to figure out what exactly got a character like his listed under Chub AI’s NSFW category.

In the end, it was clear that it was simply due to discrepancies in the platform's character categorization.

To our surprise, the initial letdown did not impede the flow of the scenario. Cyrus conversed seamlessly (so long as we kept our mature innuendos far away from him) and even proved easy to steer.

The character maintained his wary concern for the vampire he had awakened throughout the dialogue, and the scene with him ended on a warm, lighthearted note.

UX and Controls

Chub AI's message management controls are unique in that they give the user a lot of power to configure the character, chat, and memory.

In chat, you can:

• Edit all messages.
• Delete messages, although this deletion affects all messages under the selected one.
• “Continue” your messages, where the AI extends your message.
• Continue the AI’s response.
• Regenerate the AI’s response, automatically replacing the previous one, without the ability to guide its content.
• Use the “Impersonate Me” tool, which prompts the AI to write your next response.
• Make stylistic changes to the chat theme.

Chub AI also supports group chats, which can only be created from inside an individual character’s chat. Participant limits are unclear, though.

Character Library

Chub AI’s user-generated character library is quite diverse. The prominence of anime-styled character avatars reflects the community’s preferences.

• Size: Substantial
• Diversity: Anime styles are prevalent among the user community, but there are few options for users to indicate their style preferences.
• Discoverability: Sufficient. The search bar, character categories, and filtering options help you narrow down your preferences.
• Quality: Inconsistent — while the platform respects users' preferences by not displaying NSFW images, some characters (like Cyrus) are incorrectly assigned to the wrong categories.

Creating a Character

In creating your own character, Chub AI directs you to input your character details manually.

We were surprised to be greeted with a welcome dialogue containing a guide to creating characters for new users.

The basic customization options were the industry standard, with options to import your character image and definitions.

The advanced options were peculiar — two of which to note were:

• Alternate greetings — starting messages for your character to alternate between.
• Character book — defined keywords that insert specific content about your character to the AI.

The extensive character-creation tools make Chub AI appealing to users who want deep character development and world-building.

Trust, Privacy, and Data

Chub AI’s privacy policy states that the platform collects only the personal information necessary for the functioning of the site (such as usernames) and to provide services requested by the user. The policy explicitly states that users' private characters and chats will never be distributed to a third-party, unencrypted, or used for training an AI model.

The policy also lists the metadata stored and logged from queries for those who use the API outside the platform.

Chub AI warns users to read the privacy policies of any third-party sites (like APIs) that may be linked to their site, as Chub AI is not responsible for other sites.

Pricing

On their subscription page, Chub AI explicitly states that for a free trial of the platform's experience, new users receive one-time credits worth a few dozen messages.

The premium plans are:

Mercury — $5/month

• Unlimited access to Chub AI's Mistral 7B finetune and MythoMax 13B.
• 8K tokens of memory.
• Access to multimedia generation.
• Access to any future 13B (or lower) models.

Mars — $20/month

All of Mercury access, plus unlimited access to:

• Models like Asha 70B 8K and Chub AI’s beta model, Soji 671B 30K.
• Mixtral 8K 7B finetune.
• Text-to-speech.
• Multimedia generation.
• Early access to all future models.

User Sentiment

We have outlined all that we observed with Chub AI — but what sentiments do the wider Chub AI community have about the platform?

It’s clear: Chub AI is beloved among its users, with many taking to social platforms to interact and share their creations, concerns, and feedback about models and features.

The community provides constructive feedback to the platform's developers on the early access features.

Beyond the quintessential technical issues that are commonplace with platforms of such standing, there seems to be a growing trend of dissatisfied users, united by a recent pain point: subscription.

Pros and Cons

Pros

• The platform supports the use of external models from a provided list of third-party providers.
• Multiple LLM models with context models
• Extensive character creation tools.
• Strong narrative writing.
• Fast response times — 2–3 seconds on all models.
• Superior message and chat management controls.
• Consistent and well-maintained NSFW moderation.
• Users can share chats with other users.
• Users can access full chat trees with each character.
• Exciting Beta features like “Go Live”, where users can invite other users into their character chats.

Cons

• The platform only has an APK for Android devices.
• Chat and character configurations can be a steeper learning curve than on the usual RPG platforms.
• Inconsistency in character categorization
• Image generation in chat is not available

Alternatives

Admittedly, due to the lack of options in both subscription and mobile application, Chub AI, with its in-depth customization options, may not be your preference.

In that case, we recommend that you consider:

• Yodayo AI: character-driven roleplay oriented around users' preferences and control.
• Kindroid AI: provides a polished, unfiltered roleplay experience with characters that lean toward realism.
• DreamGen: offers an immersive roleplay experience — ideal for users who enjoy in-depth character customization and world-building.
• Talefy AI: provides users with guided, interactive stories with different narratives.

FAQ

Is Chub AI free?

No. Chub AI is not free. A free trial — with usage limits — is available for new users.

Does Chub AI allow NSFW?

Yes. Chub AI is an NSFW-permissive platform for users 18 and above. The platform prohibits content involving minors, real photos of real people, and illegal activities.

Does Chub AI have a mobile app?

Yes. Chub AI has a mobile app, though it is only available as an APK for Android users.

Does Chub AI support proxy and third-party models?

Yes. Chub AI supports the use of proxy and third-party models. You can set this up in your character chat settings. The platform strongly recommends using official API keys over third-party reverse proxies for security and ethical reasons.

How do I cancel my Chub AI subscription?

Canceling your subscription with Chub AI is pretty easy. Navigate to the Subscriptions page from your profile dropdown, where the platform has made accommodations for you. However, note that Chub AI does not offer refunds unless explicitly granted by the company.

How We Tested 

Test date: April 08, 2026

Platform: Web (Chrome)

Models tested: Free/Mobile, Mixtral Variant (Mars), Mistral Variant (Mercury), Soji (Mars) (Beta)

We tested Chub AI for approximately 4 hours:

• Chatted with two pre-made characters (Luther Reinhart, a mafia boss savior, and Cyrus, a wary human who awakened a vampire): approximately 25–30+ messages each
• Created 1 custom character (Sophie) with detailed definition and backstory upload; tested for approximately 30 messages.
• Conducted model comparison with Sophie (Free/Mobile and Soji) and Luther (Mixtral Variant (Mars) and Mistral Variant (Mercury))
• Conducted memory retention testing with planted details and recall checks.
• Tested voice/TTS quality with all three characters
• Tested the platform's native image generation 
• Tested steerability by attempting to redirect the conversation in multiple ways
• Tested NSFW filters permissiveness with Luther and Cyrus
• Reviewed user reviews on sites like Reddit, Discord, and other review sites from the past 6 months

We did not test: Mercury plan.

Last updated: April 04, 2026