Mammouth AI pricing, limits and the Starter plan
Mammouth AI plans, prices and usage limits, checked 6 November 2026. Work out whether Starter, Standard or Expert fits how you really use it.
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 12 and 72 a month turns into a question about your heaviest afternoon.
Everything below came off mammouth.ai/pricing and Mammouth's quota documentation. We first read them on 2 October 2026 and checked the pricing page again on 6 November 2026. Prices change, so check that date against the day you are reading this.
Plans and prices checked 6 November 2026
Plan | Billed monthly | Billed yearly (October check) | Usage vs Starter | API credits |
|---|---|---|---|---|
Starter | 12 per month | 10 per month (120 up front) | 1x | 2 USD |
Standard | 24 per month | 20 per month (240 up front) | 3x | 4 USD |
Expert | 72 per month | 60 per month (720 up front) | 10x | 10 USD |
The currency follows where you are. In October, from Europe, the page showed euros with VAT included. On 6 November, from our test location outside Europe, it showed the same numbers in US dollars excluding VAT. The monthly ladder has not moved since October. Every plan carries a "cancel anytime" note.
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 (Nano Banana, GPT, FLUX, Veo, Kling and others), Projects, voice chat and dictation. Only quota and the API credit move.
MultiChats puts the leading AI models in one app. Try it free
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 10 a month". That is the Starter rate on annual billing. Month to month, the same plan is 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:
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.
You hit it most working days, usually mid-afternoon. Standard. Three times the quota is the smallest upgrade that clears a daily interruption.
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, note that the credit is always in dollars, whatever currency your subscription is billed in.
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 6 November 2026, as on 2 October, 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 12-a-month 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 and steps you down when a model's share runs out. We meter message credits over a month: each model costs a set number of credits per reply, the model picker shows that number before you send, and the model you pick is the one that answers. When the credits are gone, you wait for the renewal or upgrade, so a monthly allowance can run out too. It just runs out in a way you can see coming.
What you get | Mammouth Starter | MultiChats Free | MultiChats Pro | MultiChats Pro+ |
|---|---|---|---|---|
Price | 12 a month, or 10 billed yearly | $0, no card | $20.99 a month or $199.99 a year | $34.99 a month or $349.99 a year |
Unit | Session quota, renews every 3 hours | 50 message credits every 24 hours | 1,500 message credits a month | 3,000 message credits a month |
When it runs low | Steps down to a lighter model | Wait for the daily reset | Credits shown in Settings; upgrade or wait | Same as Pro |
Models | Full roster on every plan | GPT-5.4 Nano, Gemini 3.1 Flash Lite, Mistral Small 4 and more | 50+ models, including GPT-6 Sol, Claude Opus 5.5, Claude Sonnet 5.5, Gemini 3.7 Flash, Grok 4.3 | Same as Pro |
Images and web search | Included in the session quota | 2 image generations a month, 2 web searches per 30 days | Both use message credits | Both use message credits; several images at once |
Apps | Web and Android | Web, iOS, Android | Web, iOS, Android, with voice calls on mobile | Same as Pro |
Ours are dollar prices before local tax. Mammouth's are shown in your local currency, sometimes with VAT included, so normalise both before you compare. The live figures are always in our pricing section.
The full feature-by-feature comparison lives on our Mammouth AI alternative post and the Mammouth alternative page. This one exists to help you pick the right Mammouth plan, including the cases where that answer is Starter. If you want to feel the difference between a session and a monthly allowance before paying anyone, open a free chat on MultiChats and run your usual afternoon through it.
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.