Mammouth AI pricing, limits and the Starter plan

To choose a plan, first check how sessions work. Mammouth meters you per session, and a session resets every three hours. Nothing here is counted in messages or in days. Once you know that, choosing between EUR 12 and EUR 72 a month turns into a question about your heaviest afternoon.

Everything below came off mammouth.ai/pricing and Mammouth's quota documentation on 2 October 2026. Prices change, so check that date against the day you are reading this.

Plans and prices checked 2 October 2026

All three include VAT and carry a "cancel anytime" note on monthly billing.

Every tier ships the identical feature list: the model roster (Claude, GPT, Gemini, Grok, Mistral, Llama, DeepSeek, Qwen, Kimi), reasoning models, web and deep research through Perplexity, image and video generation, Projects, voice chat and dictation. Only quota and the API credit move.

Step 1. Start with the billing period

Settle this before you compare tiers, because it moves every number above. Annual billing takes a sixth off the monthly rate on all three plans, as one payment up front.

Watch the headline figure. Mammouth's homepage advertises "from EUR 10/month". That is the Starter rate on annual billing. Month to month, the same plan is EUR 12. If you budgeted from the homepage and checked out on the monthly toggle, you will be 20 percent over.

Step 2. How Mammouth counts your usage

Most people skip this step. It decides your tier. There is no daily message count. Mammouth's quota documentation defines limits per session, with quotas fully renewed after 3 hours.

Six things burn through a session, in Mammouth's own list: how long your messages and the model's replies are, the size of files and images you attach, how long the conversation has grown, tools such as web search or document generation, which model you picked, and the custom instructions carried in every request.

The third has teeth. Accumulated context is resent with every message, so a long thread costs more per turn as it grows. Ten short chats and one forty-turn chat are nowhere near the same load.

Step 3. The ceiling, and what Mammouth does about it

Mammouth calls this the model threshold principle. Exhaust a model's quota inside a session and the app moves you down to a lighter model rather than blocking you. The documented example runs Claude Opus, then Sonnet, then Haiku. The pricing page words it as "If you hit a limit, a lighter model remains available."

The thresholds are not published. Mammouth documents that the swap happens and in what order, never where each threshold sits. You cannot work out in advance how many Opus turns a session buys, so if a job depends on a specific model, check which one actually answered before you act on it.

Step 4. Size the plan against your busiest session

Quota is stated as a multiple: Starter is the reference, Standard three times Starter, Expert ten times. Mammouth adds that Expert carries ten times the quota for six times the Starter price, and that 80 percent of individual users choose Starter while 50 percent of companies choose Standard.

Read your own pattern against it:

1. You hit the ceiling once or twice a month. Stay on Starter. A session renews in three hours, so the occasional wait costs an afternoon rather than a tier.

2. You hit it most working days, usually mid-afternoon. Standard. Three times the quota is the smallest upgrade that clears a daily interruption.

3. Long reasoning threads, big documents or generated media are your actual job. Expert, where the ten-times quota stops being theoretical.

Since sessions renew every three hours, what matters is what you use between 2pm and 5pm on your heaviest day. A monthly total tells you very little.

Step 5. Document limits, if you upload files

If you feed Mammouth documents, these caps matter more than the tier, since they apply to all three:

• Maximum 20 files per conversation, 100 MB combined

• Up to 100 MB for a single PDF, 30 MB for an image, 20 MB for anything else

• A long list of supported formats, audio and video included

• Maximum document content of 4,000,000 characters per request, combined across every document you attach

• Scanned PDFs that contain only images are capped at 50 pages and 20 MB

Long documents now load in full, says Mammouth. Until mid-September, long or multiple documents reached the model as an extract of the most relevant content. Mammouth's release notes dated 18 September 2026 say large documents now switch on an agentic mode so the full document and its context are loaded. With a 300-page contract, still check a summary against the pages that matter before you rely on it.

Step 6. Do the API credits change the maths?

Every plan bundles API credits and access to Mammouth Code: 2 USD on Starter, 4 USD on Standard, 10 USD on Expert. Never touched an API key? Treat that line as rounding error. If you do build things, the credit is in dollars while the subscription is billed in euros.

Step 7. Adding people

Team pricing matches the individual plans per user, and you pick which plan each seat gets. That buys an admin console for invites and licences, centralised billing, and shared projects, and it checks out like any other subscription. Enterprise is the one you cannot buy from the page: SSO, more than 100 licences, and payment by wire transfer or Chorus, quoted by sales.

Is there a free trial?

On 2 October 2026, Mammouth's pricing page, its homepage and its published documentation offer no free trial and no free tier. The terms of service say free trials "may be provided" under specific circumstances, but none is advertised. Every consumer plan starts as a paid subscription. If a third-party site advertises a trial, check when that page was written before counting on it.

Why one subscription covers so many models

How does one EUR 12 plan reach Claude, GPT, Gemini and Grok at once? Mammouth's homepage promises the leading models "at a fraction of the price" without showing the arithmetic. The mechanism sits in steps 2 and 3: a session allowance spread across the whole roster, with an automatic step down at each model's threshold.

Where we come out differently

We build MultiChats, so treat this section as an interested opinion.

The difference is the unit. Mammouth meters a session that renews every three hours. We meter messages: 50 per 24 hours on a free tier that needs no card and opens 21 of our 60 models. Pro is 20.99 USD a month or 199.99 USD a year, Pro+ is 34.99 USD a month or 349.99 USD a year, both in our pricing section. Ours are dollar prices before local tax, Mammouth's are euro prices with VAT in, so normalise both before you compare.

The full feature-by-feature comparison lives on our Mammouth AI alternative page. This one exists to help you pick the right Mammouth plan, including the cases where that answer is Starter.

How you will know you picked wrong

The invoice never moves, so it tells you nothing. Watch the step-down instead. If you keep finishing an afternoon on a lighter model than the one you opened with, the tier is too small for the work you are doing. Once a month, that is Starter behaving exactly as documented. Twice a week at the same hour, it is a plan you have outgrown.