Last week, we introduced 32K context memory for I’m All In users on GLAM Arcturus and Default, with plans to bring it to more models soon.

We’ve since added more capacity, which means we’re now ready to expand 32K support to several more models:

• Qwen Plus Character
• GLAM 5.1
• DeepSeek V4 Pro
• DeepSeek V3
• DeepSeek V3 XTC
• GLAM 4.6

That means you can now enjoy more of your favorite models with double the context, helping them keep track of more of your conversation, story, and character details.

One thing to keep in mind: double the context also means significantly more resources per generation. We’ve added more capacity to support this, but some models may still respond a little slower during peak hours over the next few weeks when dealing with traffic surge.

So if a model feels too slow at peak times, we recommend trying another model and switching back later.

Enjoy the extra memory and enhanced roleplays. 🧠💜

Hi everyone,

Over the past few years, Pixelchat has grown significantly. We've introduced more capable models, longer context, image generation, voices, memory improvements, Group Chat, Lorebooks, creator tools and much more, while keeping a large part of Pixelchat free.

Throughout that growth, our subscription pricing has stayed largely the same. However, the cost of running AI models and the GPU infrastructure behind Pixelchat has increased substantially over the last year.

To keep Pixelchat sustainable and continue investing in better models, larger context (32k is on the way) and new features, we're updating our paid subscription prices on September 8.

The most important part

If you already have an active recurring subscription, your price will not change. As long as your subscription remains active on the same tier, it will continue renewing at the price you pay today.

New monthly pricing from September 8

• Get A Taste: $5.00 → $5.95
• True Supporter: $14.95 → $19.95
• I'm All In: $24.95 → $29.95

What this means for you

Until September 8, free users can subscribe at the current price, and existing subscribers can change to another tier at current pricing. The price for the selected tier will remain locked as long as your subscription stays active.

Tier changes made after September 8 will use the new pricing.

For details about annual plans, cancelled subscriptions, non-recurring payment methods and other scenarios, please see our FAQ:

View FAQ

We know a price change isn't welcome news, and we don't take it lightly. We're announcing it in advance so everyone has time to understand their options.

Thank you for supporting Pixelchat and helping us continue improving the experience. 💜

Earlier today, between approximately 1:45 PM and 2:25 PM ET, we experienced an issue that prevented users from signing in to Pixelchat.

The issue has now been fully resolved, and sign-in is working normally again.

We apologize for the inconvenience this may have caused and appreciate your patience while we worked to restore access. 💜

Qwen Plus Character is now available again for I’m All In users.

We temporarily removed the model after identifying issues that were preventing it from running as efficiently and reliably as expected. We’ve now completed the necessary optimizations and brought it back online.

Qwen Plus Character was built for extended roleplays, with a strong focus on character consistency, long-form continuity, and staying true to the tone and personality of your character. It’s especially good at following the direction of a story over longer conversations, handling deep or brave themes, and keeping dialogue natural and in-character.

We know many of you fell in love with the model and made it your daily driver, so if you’ve been waiting to get back to it, it’s ready to use again.

Thanks for your patience while we worked on it. 💜

A lighter update this week, with a handful of fixes across web and mobile.

Recommendations fixed on Android

We fixed an issue where the Recommendations page on Android was showing the same chatbots as the homepage instead of your personalized For You recommendations.

Persona avatar generation restored

Free and Get a Taste users can once again use the Generate option for Persona avatars on Desktop and Mobile Web. Previously, only the Upload option was appearing.

Other fixes

• Fixed chatbot rating counts showing as missing or zero for some characters on the homepage.
• Fixed an issue where failed generations could appear as blank or masked responses. Users will now see a proper error message instead, making it clearer when something goes wrong.
• Fixed Anime/Semi-Realistic V2 image generation ignoring a 0 LoRA scale setting and resetting it to 0.8.
• Fixed an Internal Server Error appearing when detaching a Lorebook from a chatbot.
• Fixed an error when uploading animated WebP images as chatbot avatars.

That’s it for this week. Thanks as always for the reports and feedback. 💜

About a week ago, we released our new roleplay model, Qwen Plus Character, for I’m All In users.

After launch, we identified a few issues that were preventing the model from running as efficiently and reliably as expected, so we temporarily removed it while we investigated.

We’ve now identified what needs to be fixed, but unfortunately, the work may take a few weeks to complete.

Because of this, Qwen Plus Character will remain unavailable for the next few weeks while we make the necessary improvements.

We apologize for the inconvenience and appreciate your patience. We’ll let you know as soon as the model is ready to return.

Say hello to Qwen Plus Character, the newest model joining Pixelchat — available now for I’m All In members. 

This is a new roleplay focused model from the Qwen Family, the model is built specifically for extended roleplays. It pays close attention to the character’s persona, holds onto it across longer scenes, and follows the tone your story calls for. If you’ve ever had a model slowly drift away from who your character is supposed to be, Qwen Character Plus is designed to handle that better. 

What it does well 

• Strong character consistency — it uses the details written in the character card and stays closer to the intended personality throughout the roleplay.
• Better long-form continuity — built for extended scenes where maintaining character, tone, and story direction matters.
• Adapts to the story’s tone — whether the scene is emotional, tense, darker, or more intense, it follows the direction instead of flattening everything into the same style.
• Natural, characterful dialogue — responsive writing that feels distinct and keeps the scene moving.

  Pro tip: Qwen Plus Character performs best with a detailed and well written characters. The more personality, background, and context you give it, the more it has to work with. 

Qwen Plus Character is now available exclusively for I’m All In users. 

Give it a try and let us know how it feels in longer roleplays. 

A few useful updates went live this week, including model changes, quality-of-life improvements, and fixes across the platform.

Zeta G4 is now part of the Default model selection

Since its release, Zeta G4 has become a favorite among Free and Get a Taste users and has been performing well across the platform.

We’ve now added it as one of the models used under the Default selection. Upgrading the default experience from 8B to 26B model. 

Since the default model changes now and then, Zeta G4 is also still available separately, so you can select it directly whenever you specifically want to use it.

We’ve also made the previous default model, Lyra Spiced, available as its own option for anyone who preferred its style. It’s a fine-tuned version of Lyra that works well for lighter, general-purpose roleplays.

 Other improvements/features 

• You can now sort chats by Recent, Oldest, or Most Active.
• Restored Sort by Name on the My Chatbots page.
• Voice playback is now hidden on bot messages longer than 2,000 characters, since TTS does not support messages beyond that length.
• One of the most frustrating issues was losing a message you had spent time writing when an error occurred. Failed messages will now remain saved as a draft, so you can try sending them again instead of starting over.

Bug fixes

• Fixed an issue on Android where chatbot cards displayed every assigned tag instead of a limited number.
• Fixed an issue on Android where the Copy Message option could be cut off on longer messages.

As always, thank you for the feedback and reports. 💜

A few useful updates went out this week, including some quality-of-life improvements, an important conversation access fix, and a small model lineup change.

Favorites page now has a dedicated Groups tab

We’ve added a dedicated Groups tab to the Favorites page.

Previously, favorite groups were shown together with chatbots. You can now browse your favorite Chatbots and Groups separately, which should make the page cleaner and easier to use.

Clearer subscription info for one-time payments

Users with one-time payment subscriptions now have a Manage Subscription option that opens a read-only subscription overview page.

This page shows your plan, payment type, expiry date, and days remaining. No more guessing about when your subscription will run out or whether you need to cancel.

Conversation access fix

We fixed a major issue where users could lose access to their conversations if a character was made private or removed from the platform by the creator or moderation team.

This has now been fixed, and you should be able to access your existing conversations even if something happens to the character later.

We saw some speculation and frustration around this one, so just to be clear: this was a bug, not an intended change. Thanks to everyone who reported it and helped us track it down.

WizardLM 144B has been sunset

We also sunset WizardLM 144B this week.

The model was outdated and had very low usage, so we’ve removed it from the model lineup.

Apart from that, we’ve made a few smaller improvements and bug fixes across the platform.

As always, thank you for the reports, feedback, and patience. 💜