Release notes - Mammouth AI

September 16th 2025

⭐ 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.

August 25th 2025

⭐ 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!

August 8th 2025

⭐ GPT-5

• GPT 5 replaces GPT-4.1
• GPT-5 mini replaces GPT-4.1 mini
• GPT-5 Thinking (GPT-5 with enhanced systematic reasoning) replaces o3 & o4-mini

🦣 Each custom mammoth now has his own default model, and his own model display order, before pressing Enter

✍️ Your prompt is saved as a draft and persists when you switch conversations or refresh the page, until you send it.

August 4th 2025

⭐ New Models

New flagship models

• o3
• Magistral

New image models

• Imagen 4.0 (by Gemini)
• Grok Images

New light and fast unlimited models

• Mistral Small 32B
• Llama 4 Scout
• GPT-4.1-mini
• Grok-3-mini
• Gemini 2.5 Flash Lite

↪️ Reprompting conversation

Every user message from a conversation can now be reprompted, not only the first one. It works as a fork of the conversation.

✨ Multi-model conversation

Conversations are now specific to a brand and not a model, so you can switch from a text model, to an image model or a reasoning model in the same conversation

And also

• When a quota is reached, the conversation continues with the unlimited model of the same brand (no more starting a new conversation after quota reached)
• Grok logo has been updated from xAI logo to Grok logo
• You can delete the previous message of a conversation and therefore come back to the n-2 message and so on.
• The 404 error page (and others as well) now displays a cute mammoth

July 24th 2025

🌟 Grok 4

📂 Documents : faster, better, wider

• Scanned PDF and PDF with images are now analysed (with Mistral OCR to extract content)

• About 80 new extensions were added: more than 200 file extensions are now supported.

• Speed : Document analysis is now 3x faster

📱 Quick app install

• App can be installed from left panel menu in 1 click (from all browsers except Firefox & Safari)

🌍 New Languages

🇵🇱 Mamut mówi teraz po polsku

🇷🇴 Mamutul vorbește acum românește

🇵🇹 O mamute fala agora português

🇦🇩 El mamut parla ara català

July 3rd 2025

🌟 Midjourney v7

🧷 Mammouth API

• All subscribers get access to free token with their subscription: 2$/4$/10$ per months for Starter/Standard/Expert !

• OpenAI compatible: it can be used now in n8n, Cline, etc.. with the OpenAI plugins

• 23 text models available

• See documentation

🧮 Quota/Usage visualisation

• See your quotas at any time from your account settings

• The quota banner on the left panel can be dismissed

📂 Document & history

• While working with documents in conversation, you can know deactivate documents from past history or assistants to focus on the last document you uploaded or to stay under the 32MB limitation.

• New file extension: .doc (even if it is an old MS Office extension 😉 )

👷 Server Errors & Failed Documents

We fixed a big bunch of sporadic response errors from AI including: Server errors on Claude with uploaded images, Errors on total document upload exceeding 32MB, Errors with protected with password documents, Errors with failing API endpoints, Deleted documents that were breaking the reprompted chats with the same doc, Failed document that were blocking the conversation..

Other features

• Payment with Paypal is now accepted

• Signin with Google AND password on the same email is now working

• Reprompted conversations are now grouped together on the left panel (see gif)

June 6th 2025

🌟 Flux Kontext & DeepSeek R1

• FLUX Kontext Pro -- replace FLUX Pro & Retexture

• DeepSeek R1 28.05 -- update from DeepSeek R1

🗃️ File generation

• Generate CSV, DOCX, PDF.. with all LLMs

• Generate graphs as images with all LLMs

• See example here

📱 Mobile app revamped

• 🎙️Voice-to-text was added to dictate your prompts

• 🎨 Smoother input no more overlaps between keyboard and model buttons on new chat page as the input move upwards

• 🎨 a lot of minor UI issues (margin, borders, popups, overlaps, etc..) were fixed so the cumulated look will be more friendly

May 23th 2025

🌟 Mistral 3 and Claude 4

• Mistral Medium 3 -- replace Mistral 24.11

• Claude Sonnet 4 -- replace Claude 3.7

🧠 Smart Conversational Image

• Image models now understand when you request a modification based on the previous message

• You can see the final prompt sent to the AI behind a 🪄 icon

⚙️ Account settings -- new design of the page

• You can now delete all your chats at once
• You can now delete your account
• New sorting of user emails in Team settings (by email domain then email name)

And also..

⚡ Smoother navigation between chats -- Progressive loading of content on page with skeleton

🖋️ Last Message edition update -- When you edit your last message, you can remove the attached document

🔏 Updated privacy policy

May 9th 2025

Model Updates

🌟 Perplexity -- we added the Perplexity Sonar Pro

🌟 Gemini 2.5 Flash

And also

⚡ Multiprompting -- Simultaneous AI generation is now possible

🧠 Auto-naming of conversations

🎨 Input message preserves indentation

April 25th 2025

Model Update

🌟 GPT Images — replace DALL-E

Also,

🔍 Textual search in conversation history

Press CTRL+K or click the search icon

April 17th 2025

Model Updates

🌟 Gemini 2.5 Pro — replace Gemini 2.0

🌟 Grok 3 — replace Grok 2

🌟 GPT-4.1 — replace GPT-4o

New Photo Editing Tools

🖌️ Background remover — upload your picture and the AI will remove the background

🖌️ Quality upscaler — the AI will increase the quality of your image

Document Capacity Increased

📁 Max Input Length: 4,000,000 chars — instead of 1,000,000

📁 Max files per conversation: 20 — instead of 10

And also

📁 More extensions supported .fortran .twig .urdf .hcl .tsv .geojson .ass .tf .ps1 .scss .ino .pyw .odt

🧮 Assistant files consume less quotas: Files from assistants won't consume quotas at every prompt

And also, about product quality:

🦋 Multitasking — When you switch tabs during generation, the streaming will pursue its process

April 7th 2025

New Models

🌐🔭 Perplexity Deep Research

🌟✈️ Llama 4.0 400B Maverick — replace Llama 3.3 70B

🧮 GPT-4o-mini for documents after quota: When you reach your documents quota, you can upload more documents with GPT-4o-mini

And also, about product quality:

🦋 Unprocessed image explained — AI now provides detailed infos on why your image couldn't be processed and how you can change this. Bugs are now butterflies.

🦊 Firefox is back: UX issues happening on Firefox are fixed

🎨 AI generated hyperlinks will open in a new tab — without interrupting the streaming 😉

April 1st 2025

GPT & Gemini on Web

🌐 GPT-4o & Gemini are now routed to web search when it is relevant

🌐 Projects with mammouths — enhance and replace the assistants

🐳 DeepSeek V3-0324 — replaces the January version

💾 Conversation memory: expanded for better context

📍 Pinned Conversations: new messages pin the chat on top of the list.

🌙 Auto Dark mode — when system default is dark

🇪🇸 El Mammouth habla español

🇩🇪 Das Mammouth spricht jetzt Deutsch

March 18th 2025

MAJOR DESIGN REVAMP

🎨 Modernized User Interface — Global UI/UX revamping

📂 Projects with mammouths — enhance and replace the assistants

🎨 Modernized User Interface

✨ Model buttons — hope you'll 💖 them as much as we do

🌙 Dark mode — we changed all the dark mode colors

🎨 Sidebar color — now the chat stands out instead of the sidebar

📝 Text input:

• Filter models on category (web, text, reasoning, image)
• Drag and drop documents
• Document upload revamped with colors per extensions
• Wave black icon for vocal chat

🗨️ Chat page is revamped for large screens:

• Header: we removed the ---messages--- line and revamped the buttons
• Scroll bar: we moved it to the far right of the page
• Width: increased to max 980px instead of 768px

📂 Organize your Projects with the mammouths

• Access your mammouths in the sidebar — one-click away from your favorite mammouths

• Move your conversations from one mammouth to another

• Edit files and instructions from the prompt page

--> Assistants are dead, long live to the mammouths 🦣🦣🦣 !

And also

🖼️ Midjourney buttons: you can now click variation and upscale on all the past messages

🗃️ Errors on documents: The AI was (sometimes) ignoring your uploaded files, we fixed!

✴️ Claude logo update — replacing the Anthropic logo

March 5th 2025

Voice Mode and New Claude

🎙️ Vocal Chat — Advanced voice mode from GPT

⭐️ Claude 3.7 Sonnet — replaces Claude 3.5

And also:

🗃️ Upload files on your custom mammouths

✨ Yearly Subscriptions are available - you can switch plan in your account settings, then manage subscription, then change subscription

February 17th 2025

More context !

⭐ Max Input Size: 1,000,000 characters — instead of 300,000

And also:

🗃️ Max number of documents per chat: 10 — instead of 3

🦣 Mammouth custom instruction max size: 8,000 chars -- instead of 2,500

🇫🇷 Le Mammouth parle enfin français!

🇮🇹 Il Mammouth parla ora italiano!

February 3rd 2025

New Reasoning Models !

⭐ DeepSeek R1

⭐ OpenAI o3-mini — replace o1 & o1-mini

And also:

🧮 No more reasoning quota per day on o3 & R1 -- for now, it's like text

📣 Mammouth tips -- 6 tips to know about Mammouth

January 28th 2025

New Image Generator

⭐ Recraft V3

⭐ Recraft-svg — first svg generator

Why did we add Recraft ? Check out here

And also:

🖼️ Stable Diffusion now speaks all languages

🏗️ Unknown errors management -- more info about failed prompt

🧮 Quota details -- see when your image & document monthly quota will reset
Note: quota count is visible only when it is above 50%

∑ Nicer design for mathematic equations -- in LaTex

January 14th 2025

New Model !

🐋 DeepSeek V3

And also:

🛣 Routing by modality — Mammouth auto-determines whether to prompt text, image or web

🎨 Revamping of the new conversation page :

• on small screens, models are hidden to avoid overcrowding
• drag and drop models to sort them and choose your default one

January 7th 2025

Mammouth for Teams

⭐ Admin interface — invite team members

⭐ Centralized invoicing

And also:

🖋️ Mammouth blog — this very website 😉

🏗️ Direct access to account settings — no more email Stripe link

December 24th 2024

Model Updates

⭐ o1 — no more preview

⭐ Grok 2 — no more beta

And also:

🎨 New design for conversation page

🏗️ Adding copy output button

December 17th 2024

Model Updates

⭐ Llama 3.3 70b — run by Groq

⭐ Gemini 2.0 Flash — experimental version

And also:

🦣 Create more custom mammouths — up to 8 instead of 4

🪲 Fixed bugs happening when:

• Trying to render a mathematical Matrix in LaTex
• o1 was used with a custom mammouth
• Documents in non-roman alphabets, like Persian, were uploaded

December 10th 2024

New feature

⭐ Share your conversation with a link

And also:

🏗️ Copy pasting output preserves relevant formatting only

🎨 Sidebar can be folded

🎨 New account menu with:

• Access to your custom mammouths
• A direct link to this very release notes
• A link to Image parameters & Custom Mammouths tutorial

December 4th 2024

Image Models

🪄 Create an image from an image — with FLUX Redux

🪄 Retexture your image — with FLUX Depth

How to use image to image

And also:

📷 Upload phone camera photos and bigger images

⚡ App runs much faster when you switch conversations

🎨 Design of upload & download improved for images

November 27th 2024

New Model !

⭐ Grok 2 - new brand

November 20th 2024

Model Updates

⭐ Mistral Large 2 - November version update

⭐ GPT-o1 is now streaming

And also:

✏️ Edit your last prompt

⏹️ Stop message generation

🎨 Arrow added in the input

📁 .srt files are supported

November 15th 2024

New context size

⭐ Max input length : 300 000 characters - instead of 10k

And also:

⚖️ A quota added for long input messages

See (i) Details

🏗️ Scroll down all your conversation history in sidebar

🏗️ When opening a chat, focus is now on the last message

🏗️ Auto-scroll up the chat after new prompt

🏗️ Paste file from clipboard

🎨 New native font

October 24th 2024

Web Model

🌐 Web Search — new category

⭐ OpenPerplex — new brand

And also:

⭐ Claude 3.5 Sonnet new — model update

⭐ Stable Diffusion 3.5 Large — model update

⭐ GPT-o1 is now conversational

🏗️ Paste images from clipboard

October 19th 2024

🎨 Output Formatting

• display tables — in markdown
• titles : space & size formatting
• more space between paragraphs

∑ Display Mathematics formula — in LaTex

October 10th 2024

Files

📂 Xlsx, Xls, CSV — new extensions available for file upload

📂 Pptx, Ppt — new extensions available for file upload

🔗 URL in prompt : LLMs can screen a web page when provided with a url