French multi-AI platform
Ailix vs Mammouth AI
Two French platforms that bring several AI models together. The difference lies in what they remember about you and how you correct it.
Ailix comparisons
Mammouth, Poe, Abacus, Perplexity… For each one, what it remembers about you, what it produces and what it costs, with sources to back it up.
Ailix, online since 2024 · Comparisons updated on October 4, 2026
Each comparison gives a table, the prices we checked and the sources. Mammouth, Poe and Abacus also took the same test on an Excel budget.
French multi-AI platform
Two French platforms that bring several AI models together. The difference lies in what they remember about you and how you correct it.
Bots and models
Poe gives access to many bots and apps. Ailix 3.0 delivers a Word or Excel file directly, formulas included.
Agents and automation
Abacus hands tasks to its agents. Ailix has a second model review an analysis, with High effort from the Pro plan.
Local authorities and public sector work
Delibia relies on a body of local government documentation. Ailix produces communication materials in Word and Excel, and visuals in the Studio from the Advanced plan.
Research and execution
Two ways to search the web with several models, and what each tool remembers about you.
Research and sources
You.com has closed its chat to new subscribers. Ailix searches the web and delivers the costed recommendation in Word and Excel.
Multimedia tools
With 1min.AI, you save the context yourself. In Ailix, your conversations build up a memory that you correct.
Projects and agents
A figure changes in your file. Ailix reworks the memo and the budget from the corrected information.
API and customization
TypingMind works with your API keys. Ailix includes access to the models in the subscription.
Content and teams
Magai organizes work by client in workspaces. In Ailix, each client can have its own project, with its own memory and instructions.
Comparing answers
ChatHub shows up to six answers side by side. Ailix compares 2 to 4 models from the Advanced plan and delivers the Excel matrix.
Agents and workflows
With Flow, Magica builds procedures to repeat. Ailix digs into a one-off file and delivers the files.
Browser assistant and agent
Merlin works in the browser and on your Mac. Ailix is an online workspace that delivers files to download.
Knowledge and teamwork
TextCortex connects its agents to the knowledge of several departments. Ailix shows which information was used for an answer.
Browser and research
Sider collects information while you browse. Ailix starts from your sources and quotes to cost a campaign.
Try the platform’s name, or go back to the full list.
The thumbnails illustrate Ailix features. Fictional data.
Compare all the featuresAilix and fifteen other platforms, on ten criteria, pricing included. Choose your competitor, then open the details to find the conditions and the sources.
Scroll the table horizontally to see the other platforms.
| Criteria Checked on October 4, 2026 | AilixExplore Ailix | Mammouth AIRead the comparison | DelibiaRead the comparison | PoeRead the comparison | Abacus.AIRead the comparison | 1min.AIRead the comparison | You.comRead the comparison | GensparkRead the comparison | PerplexityRead the comparison | TypingMindRead the comparison | MagaiRead the comparison | ChatHubRead the comparison | Magica (ex-Galaxy.ai)Read the comparison | MerlinRead the comparison | TextCortexRead the comparison | SiderRead the comparison |
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Personal memoryWhat is remembered and what you can edit. | Editable memory, read by every model A memory that every model in the chat reads, built up over your conversations from the Starter plan. In My memory (Notebook view), you correct, pause, export or erase what has been remembered. See the example | Extended memory within the conversation | Not established in the sources consulted Not established in the sources consulted. | Optional memory, shared across several bots Optional memory, updated daily, that you can view and delete. Shared by most official conversation bots. [1] | Persistent memory in Pro Personal Agents Personal Agents on Pro with persistent memory. Detailed correction controls not established here. [3] | Saved memory, editable and deletable Manual saving of context, with editing and deletion. Information reused across conversations on compatible models. [2] | Personalization that can be enabled in agents Personalization that can be turned on in agents. Detailed controls for a personal memory not established in this research. [3] | SecondBrain and connected personal sources SecondBrain links activities, meetings and connected sources to a personal memory. Connections can be removed. [3] | Memory and Brain with separate controls Search memory that you can view, delete and turn off. Brain, a separate feature, serves Computer and is reserved for Max subscribers. [2] | Profile and memory through an MCP server | Personal Context with explicit instructions Personal Context keeps explicit information and instructions, applied to conversations across teams and workspaces. [1] | Not established in the sources consulted Not established in the sources consulted. | Facts you can view, edit and delete Learned facts that you can view, edit or delete (Settings, Memory). [3] | Memory announced in Merlin Chat Memory announced in Merlin Chat. Detailed controls for viewing and correcting it not established here. [5] | Persistent memories added to personal context Memory tool that persists across conversations, with manual creation of information about personal context. [3] | Not established in the sources consulted Not established in the sources consulted. |
| Project contextKeeping information separate or shared between folders. | General memory or a project’s own memory Projects with files and instructions. A project uses your general memory or a memory of its own. | Custom Mammouths with documents and sharing | Personal spaces and shared assistants | Not established in the sources consulted Not established in the sources consulted. | Projects shareable within the organization Projects that can be shared with members invited to the organization. Global instructions that apply to conversations. [4] | Instructions kept, isolation not established Memory can keep project or brand instructions. Separation between projects not established here. [2] | Agents with instructions, files and sources Custom agents with instructions, files and sources. Memory isolated per project not established here. [3] | Hub shares files, instructions and history A Hub shares files, instructions and history across its projects. Members can pick up shared work. [4] | Shared Brain and optional personal memory | Instructions and documents in projects Project folders pass instructions and documents to conversations, with a default model or agent. [4] | Workspace and agent context | Not established in the sources consulted Not established in the sources consulted. | Shared or Isolated memory, set at creation Shared projects with global and project memory, or Isolated projects with project memory only. The choice is set at creation. [4] | Projects, history and brand voice | Agent-specific instructions and knowledge Agents with their own instructions and knowledge. Memory isolation per project not established in the sources consulted. [7] | Not established in the sources consulted Not established in the sources consulted. |
| Reference documentsFiles, search in your documents and knowledge bases. | Documents and knowledge with citations Attached files and selected knowledge bases ground answers in your documents, with citations. | Attachments and Mammouth documents | Domain corpus and internal documents | Files attached in conversations Files attached to conversations. A persistent knowledge base is not established in the sources consulted. [1] | Documents, spreadsheets and internal sources Analysis of attached documents and spreadsheets. Agents configurable on internal data, with Drive, Slack or Confluence connections. [4] | Documents and web search in conversation | Search across public and private sources | AI Drive, SecondBrain and shared documents | Files and tools kept per Project Projects keep files, instructions and connected tools in a persistent context, shared according to permissions. [5] | Documents and RAG knowledge depending on the plan | Lasting knowledge and attachments | Questions about attached documents Attachments to query documents, including PDF, DOC and TXT, within a conversation. [5] | Project files and a reusable Library | Attached documents and local files on Mac | Knowledge bases with cited excerpts | Wisebase and cited Deep Research reports |
| Models and routingManual choice, automatic selection or coordination of models. | More than 100 models, automatic or manual choice GPT, Claude, Gemini and more than 100 models in one chat. Ailix 3.0 chooses the model and the tools, or you choose yourself. Compare shows 2 to 4 answers side by side, from the Advanced plan. Learn more | Manual choice and automatic routing | Announced orchestration, Mistral by default The vendor announces orchestration of several models, with Mistral models used by default. [5] | Bots of your choice and the automatic Assistant | Automatic RouteLLM or a custom router Manual selection, automatic RouteLLM routing and the option to build your own model router. [1] | Several models in Multi AI Chat Choice of several models in Multi AI Chat, with a unified conversation history. [1] | Auto, Express, ARI or a chosen model Agents can use Auto, Express, ARI or a chosen model, with their own instructions. [3] | Super Agent and models of your choice | Model choice and Computer coordination | Providers and models per agent | Manual models or Auto selection | Up to six models in parallel | Automatic routing or direct model choice | Merlin Magic picks the model and mode Merlin Magic documents automatic selection of the model and mode. Usage varies by tool. [4] | Model chosen per chat or agent Model choice in the chat and a configurable default model for each agent. [7] | Sider Fusion picks the model Sider Fusion chooses the model automatically. [5] |
| Effort and second reviewDepth of processing and answers weighed against each other. | Four effort levels, second opinion from Pro Fast and Standard from Starter. In High, from the Pro plan, a second model from another provider reviews the work before the answer. Max from the Max plan. | Same request sent to another model A request can be sent again to another model. Distinct effort levels not established in the sources consulted. [1] | Not established in the sources consulted Not established in the sources consulted. | Several bots in the same thread A single request can call on several bots at once and bring their answers together in the same thread. [2] | Adjustable effort levels in Agent Agent offers xLow, High, xHigh and Max, among others. Demanding tasks use more credits. [2] | Model answers side by side The same request is sent to several models, with the answers shown side by side. [1] | Configurable search depth and reasoning | Standard or Ultra depending on the product Standard and Ultra levels depending on the product. Usage varies with the model, the context and the steps. [2] | Model Council weighs several models Model Council queries several models separately, then summarizes their conclusions and disagreements. On the web with Max or Enterprise Max, in all paid plans through Computer. [3] | Parallel answers and agent chains | Not established in the sources consulted Not established in the sources consulted. | Compared answers, then an optional summary | More processing capacity in Max mode Auto and Max drive the processing. Max uses more powerful models, according to the provider’s documentation. [2] | Not established in the sources consulted Not established in the sources consulted. | Low, Medium, High and Thinking where supported Low, Medium and High levels, with Thinking on compatible models and default settings that admins can manage. [4] | Not established in the sources consulted Not established in the sources consulted. |
| Office filesDocuments, spreadsheets and presentations to rework. | Word and Excel, presentations in the Studio Downloadable Word and Excel files, with formulas, from Starter. Presentations in the Studio, from the Advanced plan. | Documents created and edited by an agent Document generation and editing with a virtual computer, specialized tools and sub-agents. [2] | Administrative writing, exports not established | Formats vary by bot or Canvas | Documents, spreadsheets, PowerPoint and reports | Presentations, Word and Excel exports not established Presentation generator announced. Word and Excel export formats not established in the sources consulted. [3] | Cited research reports and PDF ARI produces research reports with citations and PDF export, from public or private sources. [5] | Docs, Sheets and Slides with editing | DOCX, XLSX and PPTX with versions Creation of DOCX, XLSX and PPTX files, with preview, versions and download. Changes are requested in the conversation. [4] | Sandbox files, Canvas editing and Artifacts | Editable Canvas, DOCX and PDF export | Not established in the sources consulted Not established in the sources consulted. | Editable Word, spreadsheets, presentations and PDF | Documents and presentations through Merlin Agent Merlin Agent announces document and presentation production, as well as work on local files. [2] | Writing and analysis, exports not established Document drafting and analysis. DOCX, XLSX and presentation exports not established in the sources consulted. [7] | Documents, presentations and data analysis Document and presentation creation announced, with data analysis and production tools. [2] |
| Visual creationImages, videos and creative tools. | Images and videos in the Studio, from Advanced Images, edits and videos in the Studio, from the Advanced plan, following the folder’s instructions and brand guidelines. | Images, videos, music and editing | Not established in the sources consulted Not established in the sources consulted. | Images, videos, audio and music Bots and models for image, video, audio and music generation. [1] | Studio for images, videos, music and voice Studio for images, video, music and voice. Editing, ad and avatar tools announced. [1] | Images, audio and videos on credits | Not established in the sources consulted Not established in the sources consulted. | Image and video generation Image and video generation tools, with a choice of model and settings that affect usage. [2] | Images and videos depending on the plan | Image creation and editing through plugins Image generation and editing through plugins, including GPT Image Editor and Stable Diffusion, to be configured. [5] | Images and videos in Media | Images, videos and image editing | Images, videos and audio with Flow | Images and videos through Merlin Agent Merlin Agent announces image and video generation, in addition to documents and presentations. [2] | Images with a configurable model per agent | Images, videos and creative tools |
| Agents and actionsTools, connectors and task execution. | Ready-to-use tools, custom agents Ready-to-use tools from Starter (report, translation, transcription…). Custom agents from the Advanced plan. Ailix 3.0 runs web search, calculation and files as the request requires. Learn more | Sub-agents, connectors and MCP servers Agentic mode, sub-agents and Microsoft 365, GitHub, Notion and Atlassian connectors. Custom MCP servers. [2] | Assistants for local government roles | Canvas apps that call other bots App Creator generates interactive Canvas apps that can call other Poe bots. [3] | Agents, integrations and scheduled tasks | Not established in the sources consulted Not established in the sources consulted. | Customizable research agents with tools Customizable research agents, with instructions, files, sources and personalization settings. [3] | Super Agent and reusable Skills Super Agent plans and executes with a browser, files and tools. Skills make some workflows reusable. [1] | Computer runs and schedules tasks Computer runs tasks with tools, works in parallel and can continue or schedule tasks. [1] | Agents, workflows and MCP tools | Agents, integrations and scheduled prompts | Not established in the sources consulted Not established in the sources consulted. | Connected agent and visual Flow workflows | Mac agent and browser assistant | Workflows, scheduled tasks and connected tools Agents connected to company knowledge and tools, visual workflows, scheduled tasks and communication integrations. [2] | Agent that acts in the browser Browser agent that can act on open sites and sessions, reading the context of pages. [1] |
| TeamworkSharing, members and administration. | Invitations and roles in teams Team workspaces with invitations and owner, administrator and member roles. Only the people in charge manage members. | Sharing, permissions and provider restrictions Roles and permissions, shared Mammouths, provider restrictions. Documented SSO (Entra ID, Okta, Keycloak, SAML 2.0), tied to the Enterprise plan from 100 licenses. [4] | Permissions, support and team training | Group chats and shared points Group chats with invitations and administration. Some plans allow a points budget to be shared. [1] | Shared members, projects and conversations Invitations to an organization, sharing of conversations and projects, with team subscription management. [4] | Collaboration with Business or Enterprise Sharing and collaboration offered in the Business and Enterprise plans, with different team capabilities. [5] | Max plan, existing subscribers only For existing subscribers, the Max plan announces collaboration with a team of 25 people. Detailed permissions not established here. [2] | Managed Team and Enterprise governance | Shared projects and Enterprise controls | Central keys, rights depending on the Teams plan | Workspaces and monthly limits per member The Team holds the subscription and can set monthly limits per member. Workspaces manage members and content permissions. [2] | Team plans on request, per seat Team plans on request with per-seat discounts. The exact terms are set with the provider. [6] | Workspaces, roles and central billing Team workspace, admin dashboard and centralized billing. Team from five users. [10] | Members, usage tracking and team bots Member and organization management, usage dashboard, brand voice and custom bots for teams. [3] | Admin control of agent rights and spending | Team plan per seat, one administrator Team plan per seat, billed monthly. A single administrator, the owner. [2] |
| Price and usageDifferent currencies and scopes. Credits are not equivalent. | Starter: €15 excl. VAT / month 20 monthly credits. Advanced team: €20 excl. VAT / license / month, 30 credits. A Free plan is also available. Access to tools and models depends on the plan. | Starter: €10 excl. VAT (€12 incl. VAT) / month Quotas per session, renewed every 3 hours. Switches to a lighter model at the threshold. Plan and conditionsStarter: €10 excl. VAT (€12 incl. VAT) a month, Standard €24 incl. VAT, Expert €72 incl. VAT, billed monthly or annually. Usage depends on quotas renewed in three-hour sessions. Recorded on October 4, 2026 on Mammouth’s pricing page, which shows prices including VAT. Mammouth quotas and Ailix credits do not convert into each other. Pricing and sources of the comparison | Free license for small municipalities, otherwise a quote The quote depends on access, support and the terms agreed with the local authority. Plan and conditionsFree license for town hall clerks (Small municipalities plan). Essentiel and Personnalisé on quote, with a 14-day trial. The organization-wide plan can include licenses for staff and elected officials, plus support. The Delibia quote covers an entire local authority. We do not derive a per-license price from it or compare its price with Ailix. Pricing and sources of the comparison | From $4.99 / month Points budget depending on the plan and the bot. Top-ups available, taxes depending on the subscription. Plan and conditionsPoe advertises subscriptions starting at $4.99 a month. The points used vary with the bot and the request, and the amount included depends on the plan. Additional points can be purchased. Checked on October 4, 2026. Taxes depend on the subscription. Poe points and Ailix credits do not convert into each other. Pricing and sources of the comparison | Basic: $10 / person / month Basic caps Agent and Studio tasks. Credit limits still apply on Pro. Plan and conditionsChatLLM Basic is listed at $10 per user per month with 20,000 credits. Pro costs $20 in total with 30,000 credits. Basic limits Agent tasks and Studio conversations. Pro removes these conversation caps, within the limit of available credits. Checked on October 4, 2026. A first-month promotion can change the price and the credits included. Taxes depend on the payment. Pricing and sources of the comparison | Pro: $6.50 / month Credit allowance depending on the plan. The units are not equivalent to Ailix credits. Plan and conditions1min.AI lists Pro at $6.50 a month with 1 million bonus monthly credits, Business at $10 with 2 million, and Enterprise at $7 per member per month. A 20% discount is advertised for annual billing, and unused credits roll over. Checked on October 4, 2026, taxes depend on the payment. These credits do not convert into Ailix credits. We did not include lifetime deals sold by resellers. Pricing and sources of the comparison | Chat closed to new subscribers, API billed by usage Since August 4, 2026, no new chat subscriptions. Existing subscribers keep their plan. Plan and conditionsSince August 4, 2026, You.com no longer opens new subscriptions to its chat (Pro, Max). Existing subscribers keep it. The current public offer is an API billed by usage. Checked on October 4, 2026. A subscriber who cancels can no longer resubscribe. The chat features described on this page (agents, source control, ARI) concern existing subscribers. Pricing and sources of the comparison | Team: $30 / seat / month, 2 seats minimum Team price with monthly billing. Credits and top-ups depending on the tools. Taxes at checkout. Plan and conditionsPlus costs $24.99 a month, or $19.99 billed annually, and Pro $249.99, or $199.99 billed annually, with different credits and storage. Team is documented at $30 per seat per month with monthly billing, with a minimum of two seats. The tools used and top-ups can change the total cost. Checked on October 4, 2026. Reduced prices and some benefits depend on the billing cycle and sometimes on how long you have been subscribed. Taxes and final total at checkout. Pricing and sources of the comparison | Pro: $20 / month, no recurring monthly Computer credits The Pro Computer bonus is temporary and one-time. Regular use: extra credits or another plan. Plan and conditionsPerplexity Pro is listed at $20/month. On Pro, Computer does not provide recurring monthly credits. The 4,000-credit bonus is one-time and valid for 30 days. Max is offered at $200/month or $2,000/year on the web and includes 10,000 monthly Computer credits. US dollars, checked on October 4, 2026. Taxes and local price at checkout. Promotional bonuses are distinct from monthly credits. Ask searches do not use Computer credits. Pricing and sources of the comparison | License from $39 one-time + API usage Personal license and Teams are separate. APIs are paid separately from ChatGPT or Claude subscriptions. Plan and conditionsTypingMind offers one-time-purchase personal licenses: Standard $39, Extended $79 and Premium shown on promotion at $99. API usage is still billed separately. Teams starts at $99/month with monthly billing, with five seats included. Growth and Professional are listed at $199 and $299/month with monthly billing, excluding API usage. Prices checked on October 4, 2026. Check the promotion and taxes at checkout. ChatGPT or Claude subscriptions do not cover the use of their APIs. The personal license and Teams are separate offers. Pricing and sources of the comparison | Standard: $20 / month Usage allowance and available models depending on the plan. Top-ups available. Taxes and seats at signup. Plan and conditionsMagai lists Standard at $20/month, Pro at $40/month and Ultra at $200/month. The detailed table distinguishes standard and premium models. Usage draws on an allowance that varies by model and operation, and top-ups are offered. Monthly prices checked on October 4, 2026. Taxes and seats depend on the subscription. Available models and usage limits depend on the chosen plan. Pricing and sources of the comparison | Pro: $19 / month, billed monthly Annual prices require annual payment. Images and videos have their own limits. Plan and conditionsChatHub Pro is listed at $19/month or $14.99/month billed annually. Unlimited is listed at $39/month or $24.99/month billed annually. Pro includes request and document quotas. Unlimited advertises these uses with no quota, with a separate allowance for media. US dollars, checked on October 4, 2026. Annual prices assume annual payment. Taxes and media units at checkout. Pricing and sources of the comparison | Team: $15 / person / month, 5 people minimum Team price, additional taxes not included. Individual plans have separate terms. Plan and conditionsMagica offers several individual plans, to be checked in its current offer. The documented Team price is $15 per user per month, or $96 per user per year (that is, $8 a month), with a minimum of five users. US dollars, taxes extra according to the terms, checked on October 4, 2026. We do not quote the individual price, as we could not confirm its exact terms. Pricing and sources of the comparison | Free and paid plans Usage varies with the model and the feature. Top-ups and team-specific terms. Plan and conditionsMerlin offers a free plan and a Pro plan at $228 a year. Teams costs $19 per seat per month billed monthly, or $15 billed annually, with a two-seat minimum. Deductions vary by model and feature, and top-ups are available. Plans checked on October 4, 2026. The applicable price depends on the plan, the billing period and the number of users. We do not compare cost at equal usage here. Pricing and sources of the comparison | Premium: $29.99 excl. tax / person / month Monthly billing, compute budget included. Additional usage can be enabled. Plan and conditionsTextCortex lists Premium at $29.99 per user per month excl. tax, billed monthly, with a compute budget. The documentation provides for a fallback to standard models when this budget runs out, and an optional Extra Usage that admins can manage. Checked on October 4, 2026. The cost depends on the plan and on any extra usage turned on. The TextCortex compute budget does not convert into Ailix credits. Pricing and sources of the comparison | Nano to Max: $100 to $1,000 / year Annual billing shown. Credit volumes depending on the plan. Team plan per seat, billed monthly. Plan and conditionsSider presents the Nano, Lite, Pro and Max plans, from $100 to $1,000 a year billed annually ($8.33 to $83.33 a month), with different credit volumes. Credits are deducted based on actual usage, and these rules replace the former Basic, Advanced and Elite grids. Plans checked on October 4, 2026. The applicable amount is shown in the chosen plan, and each feature uses credits. Pricing and sources of the comparison |
The details give the conditions and the references [1], [2]… of each comparison. “Not established” means that the sources consulted do not allow a conclusion about this feature.
Read the detailed comparisonsWhen a preference or a deadline changes, you find the information in My memory (Notebook view) and you correct it.
01 · View the memory
Nonprofit · Sample personal sheet
A cultural nonprofit based in Nantes.
02 · Correct a fact
The recorded response time is 48 h. You can update this information without erasing the sheet.
In the Notebook, you can edit a fact on the sheet. Here, the response time would go from 48 to 24 hours.
03 · See the context used
“Draft an email about our response times.”
Note available in the sheet: a response within 48 h.
The Details panel shows the memory folders actually used for an answer.
“Our next workshop is called Lueur.”
Atelier Lanterne Folder updated
Next workshop: Lueur
A separate example of a new fact remembered. This receipt describes an addition, not the folders used to answer.
Illustration of how it works, with fictional data from Atelier Lanterne. This walkthrough does not start any conversation.
These pages are written by Ailix. They rely on each platform’s official documentation, supplemented by our own tests when we have an account. We distinguish documented features from the results we actually observed.
We also point out the cases where we would rather recommend the competitor. If a feature does not appear in its documentation, we do not conclude that it is missing. The sources and the limits of our tests are given on each page.
See Ailix on a real case