Mammouth AI Release Notes

⭐ New Models
Video Generation - Video generation is now available in Mammouth with 4 models: Veo 3.1 fast, Sora 2, Grok Video & Kling 2.5
Grok 4.20 - The new Grok model is available with updated web search including specific x.com search.
Light models updates:
• Claude Haiku 4.5 is added
• GPT 5.4 nano replaces GPT 5 mini
• Mistral Small 4 replaces Mistral Small 2
👥 New Features for Teams and Companies
SSO Integration - We now support SSO integration with Microsoft Entra, Keycloak, Okta, and more. It facilitates deployment of Mammouth across companies or universities.
Multi-licensing - Your team can now mix plans (Starter/Standard/Expert) in the same subscription to align with various user needs.
More details on teams:
📊 New Quota Policy & Increased Context Capacities
Updated quota policy:
• The monthly quotas on images and documents have been removed
• The fixed 50 messages quota has been removed too
• All limitations are now calculated in the 3-hours session quota that depends on the real consumption of AI (number of messages, length of messages, model selected, reasoning, web search)
• After-quota model switch has been improved to be more progressive and less frustrating
Increased context on documents - We added an option in user settings to increase the context on documents (for specific needs on very long documents).
Increased memory - We added an option in settings to increase the context size of the conversation memory that can be sent at every message.
More details:
And also
• Text to Speech (with Voxtral) on output message so you can let the AI read its answer out loud
• New file extensions supported : .ods, .odp, .odt
• Revamped design on chat messages to remove borders and be more mobile friendly
• Transactional emails templates (email verification links, forgotten passwords, team invite etc.) were fully redesigned
• Conversation filters per mammouth and per date were added:
• You can bulk select/move/delete your conversations older than 3 months
• You can bulk select/move/delete your conversations from a custom mammouth

New Models

Gemini 3.1 Pro

Recraft v4

Shared Mammouths

Mammouths can now be shared via public link, email invitation or within your Team.

And also

• When the LLM output is still streaming, you can now copy code blocks or table or open external links.
• We now accept all text file extensions. This is especially useful for coders using niche file formats (such as .csproj or .mq4).
• Models were added in the API (GLM-5, Gemini 3.1 Pro, Claude Sonnet 4.6, gemini 2.5 flash lite). Additionally, 13 models were deprecated to keep the list clean and up to date.
• The delete chats feature was improved — the delete operation now runs in the background, so you can keep working without waiting on a frozen screen.

New Models

Claude Sonnet 4.6 - just released!

Grok Imagine - replaces Grok 2 image

Model Updates

• GPT Image - updated for better addressing text in images in large prompts

New Features

• Copy while streaming - you can now copy the AI's output text while it is still live streaming
• Document generation with reasoning models - document generation now works when you chat with reasoning models such as Gemini Pro or Claude Opus

Settings & Input

• New settings page - the settings page was fully revamped with a modern design
• Switch ENTER key behavior - choose which key will send your prompt: ENTER, ALT+ENTER, or CTRL+ENTER in your settings

New Mammouth Accessories

Check out the new temporary mammouth accessories ⛷️

New Models

Claude Opus 4.6 and Kimi K2.5

Mammouth Code !

Check out the new Mammouth Code documentation for details.

Web Updates

• Image search : search queries can now display images from the web
• Simplification : the web category line was merged with text models and OpenPerplex was removed. You can still force select the web category.
• Perplexity : now has a new display with snippets

Other

• Reprompting can now ricochet across multiple conversations
• When you copy an output message, it will paste with the applied markdown formatting (more convenient for emails, docs, etc.)

New Model

Mistral Large 3 - replaces Mistral Medium

Model Updates

• Gemini - improved stability and reliability
• Gemini - now supports YouTube video link screening

Conversation Management

• Batch operations - delete or move multiple conversations at once
• Enhanced search - find conversations by model or mammouth name

Mammouths

• Unlimited mammouths - no more 15-mammouth limit
• Favorite mammouths - mark your most-used mammouths as favorites
• Smart ordering - the last used mammouth automatically moves to the top of the list
• Improved sidebar navigation - smoother browsing through your mammouth collection

New Mammouth Accessories

Seasonal accessories - collect limited-time accessories throughout the year! How many will you catch?

Mobile App (Chrome users)

Reinstall for new features - update your app to access native capabilities:

• Quick shortcuts - press and hold the app icon to reveal action shortcuts
• Enhanced sharing - seamlessly share photos and files directly in Mammouth

And also

• Dynamic browser tabs - conversation names and mammouth icons now appear in browser tabs
• Nano Banana API - added as the first image model in the Mammouth API
• GPT Image improvements - enhanced image generation quality and capabilities
• VAT-inclusive pricing - all EU countries now see prices including VAT

⭐ New Premium Model

Claude Opus 4.5 - The most powerful model from Anthropic joins Mammouth

🔄 Model Updates

• GPT 5.2 - replaces GPT 5.1
• Gemini 3 Flash - replaces Gemini 2.5 Flash
• GPT Image 1.5 - replaces GPT Image 1

🌟 New Brands

Kimi - New brand with two models:

• Kimi K2
• Kimi K2 Thinking

Qwen - New brand with:

• Qwen 3 Coder

📂 Enhanced Document Context

8x more context extraction - The extracted context from documents in your mammouths has been multiplied by 8, now matching the context capacity of conversations.

⚙️ Advanced Settings Control

• Deactivate API settings - Turn off API access directly from the menu
• Brand management - Deactivate specific model brands for you and your team from permission settings

🌎 New Language Support

🇳🇱 De Mammoet spreekt nu Nederlands!

New Design : Mammouth gets a makeover !

Discover our new visual identity, enhanced customization options, and a redesigned user interface for an even better experience :

New Logo & Customizable Mammouths

• New logo : While preserving Mammouth's iconic spirit and its original icon, we've modernized the shape and colors for a more dynamic and contemporary look !
• Custom Mammouths Experience : Personalize your mammouth with a wider range of colors and even more props to make it truly unique !

Improved user experience (especially on mobile)

• Larger buttons : for easier and more intuitive navigation.
• New color palette : a refreshed and harmonious interface.
• Fully responsive mobile : perfect fluidity across all your devices.
• New dark mode : a more coherent and pleasant display for your eyes.

Flux 2 Pro - replaces Flux Kontext Pro

DeepSeek v3.2 - replaces DeepSeek v3.1

And also

Pin your conversations to keep them on top of the list

File generation we improved the file generation system

Reprompting improvements

• Perplexity was added to the list reprompting list
• Upon reprompting on Grok, DeepSeek or Mistral, the web search will automatically activate if relevant

API update: Claude Opus 4.5, GPT 5.1, GPT 5.2, Gemini 3 Pro, Perplexity Sonar Pro, Perplexity Deep Research, Grok Code Fast, Claude 4.5 Haiku, Qwen3 Coder, Kimi K2, Mistral Large 3 and DeepSeek v3.2 were added to the API

Team Roles

if you use mammouth in your company, you can now:

• Assign roles : owner, admin (team management to activate/deactivate members), members
• Transfer ownership: the owner can designate a successor
• Administrative account: the owner of the plan don't need to use a licence for himself
• Batch invitations: invite multiple email addresses at once

And also

Grok 4.1 fast

replaces Grok 4

Export your conversations

export available in JSON format

Import your conversations from chatGPT

you can now

A product detail

if you ask a text model like Mistral or DeepSeek to generate an image, it will create a smart link to quickly reprompt your query to an image model.

Fixed an occasional error on GPT-Image when sending emojis in prompts.

GPT 5.1

⭐ GPT 5.1 - replaces GPT 5 and GPT-5 Thinking. The model is faster while keeping up with GPT 5 performance.

Grok 4 fast

⭐ Grok 4 fast - we updated Grok 4 with a faster text model, the new light model and a reasoning model.

Nano Banana

🍌 Nano Banana - we updated Nano Banana to the post-preview version. It now handles aspect ratio parameters.

Auto-activation of web search

🌐 Auto-activation of web search -- Mistral, Grok and DeepSeek now search the web automatically when relevant. You can still deactivate or force the option by clicking on the web icon while prompting.

And also

✉️ The account email can now be changed in the settings.

⏬ A shortcut button to scroll down to the end of the conversation was added.

🪛 A bug was fixed on Safari last released version that prevented AIs from reading documents containing French accents in the title. Additionally, we fixed an issue on some code imports that were inadequately restituted by the AI when uploaded as documents.

🧷 Embeddings are now available in the Mammouth API.

⭐ Recraft Vectorize (PNG to SVG) ⭐

Recraft can now vectorize your PNG or JPEG images into .svg format. It is the only model that can do so.

🐳 DeepSeek v3.1 Terminus

we updated the DeepSeek v3.1 to its latest version

🔉 Audio-to-text is now available

you can upload and analyze audio files with Gemini

🔉 Video-to-text is now available

you can upload and analyze small video files (less than 20MB) with Gemini

And also

💭 Thinking mode is now less invasive with a smoother design. Also, the web search mode from GPT is now displayed.

🦊 The app is now available for Firefox users on Windows.

🪛 Fixed accidental prompt sending when Shift is released before Enter (when using Shift + Enter).

➗ A few LaTeX formula fix were deployed for a nicer display of math equations.

📋 Loading history chats in the sidebar now works on large screens too.

⭐ Claude Sonnet 4.5 ⭐

It replaces Claude Sonnet 4. See the release note from Anthropic here

🧠 Memory of the conversation

We've unleashed conversational memory. Your model now has access to the entire message history, no matter how long the conversation.

🎨 Design of the AI answer

The output formatting of the AI responses has been improved. The spaces between paragraphs and titles are now more adjusted. The quotes, the tags and the generated tables have a better color design.

⛵ Bye bye Midjourney

Last week, our Midjourney provider discontinued service. We haven’t found a suitable replacement, as Midjourney still doesn’t offer an official API. Although it ranks 4th in our image-generation usage (behind Nano Banana, GPT Image, and FLUX) and 5th in performance benchmarks, it still holds a special place in our hearts. We might reproduce its four‑image grid experience with other models. We hope Midjourney eventually launches a developer-friendly API. Until then, we'll sail our separate ways.

And also

• GPT Image is now fully conversational (like Nano-Banana).
• Bug fixing: some models were occasionaly trying and failing to execute code while not requested during document analysis or graph analysis. It's fixed.
• Mammouth API: Claude Sonnet 4.5, Grok-4, Grok-4-fast, DeepSeek-3.1-Terminus and DeepSeek-3.2-exp are now supported.

New models

• Nano Banana (aka Gemini 2.5 Flash Image): can edit images and handle conversations.
• DeepSeek v3.1 (replace DeepSeek v3)

Start a conversation with light models

You can start a new conversation with any light model even if you've not reached your quota.

API services for your automations

• Buy additional credits in the api settings.
• You can also buy credits without having a subscription.
• GPT-5, GPT-5-mini, GPT-5-nano, Claude 4.1 Opus, Mistral 3.1, DeepSeek v3.1 are now supported.

And also

• Improved handling of multiple documents in a conversation: short documents were sometimes overshadowed by long ones and the AI missed them; that's no longer the case.
• Document generation : some models occasionally refused to generate a file (pdf, docx, xls); this is now fixed.
• The pricing page was updated to highlight the Team plan and provide up-to-date information about the growing Mammouth offering (API credits, Perplexity, Vocal mode, etc.).
• We updated our Terms of Service.
• GPT-5 now comes with:
  • better text formatting
  • functional math equation rendering.
  • the thinking process is now visible.

New model: Mistral Medium 3.1

Long Context & Documents

We've significantly improved the handling of long contexts (long input & documents): more context capacity, better content extraction, improved management of long code context, faster performance, fewer errors etc...

UX Improvements

• The code-copy button is now floating for easier access
• Copying and pasting emails or HTML content no longer adds unwanted line breaks
• Refined mobile navigation for a smoother experience
• More custom mammouths are now visible in the sidebar
• Max mammouth count increased from 12 to 16.

Web Access for all AIs

We've expanded web capabilities to Mistral, Claude, DeepSeek, Grok and Llama!