Fair Usage Policy

Model rate limits & usage guidelines for trial and professional plan users

This Fair Usage Policy governs the use of AI models available through the DeepMask platform. It is designed to ensure equitable access, platform reliability, and a consistently high-quality experience for every user. DeepMask provides access to 18+ leading AI models from providers including OpenAI, Anthropic, Google, Meta, Mistral, xAI, Moonshot AI, and more. Because these models have varying computational costs and demand profiles, we apply tiered rate limits to prevent any single user from monopolizing capacity and impacting the experience for others.

Our goal is simple: every DeepMask user should be able to access any available model, at any time, without degradation. This policy makes that possible.

Model Categories & Rate Limits

Model Categories & Rate Limits

DeepMask organizes its AI models into four categories based on computational cost and demand. Each category has its own rate limit, measured in messages per 3-hour rolling window. These limits apply per user, per category.

Rate Limits by Plan

Fast

Fast

Balanced

Advanced

Premium

Trial (Free)

Pro (Paid)

50/3 hrs

30/3 hrs

10/3 hrs

5/3 hrs

250/3 hrs

100/3 hrs

50/3 hrs

20/3 hrs

Fast

Fast

Balanced

Advanced

Premium

Trial (Free)

Pro (Paid)

50/3 hrs

30/3 hrs

10/3 hrs

5/3 hrs

250/3 hrs

100/3 hrs

50/3 hrs

20/3 hrs

When a user reaches the limit for a given category, they are temporarily unable to send new messages to models in that category. They can immediately switch to any model in a different category and continue working. Limits reset on a rolling 3-hour window — so access is restored automatically as time passes.

Most users never reach these limits. They exist as a safeguard, not a barrier. If you regularly hit limits, consider upgrading to the Professional plan for significantly higher allowances.

Model Categories Explained

Model Categories Explained

  1. Fast Models

Lightweight, high-speed models optimized for quick queries, simple tasks, drafting, and high-volume usage. These models offer the lowest cost per token and the highest rate limits — ideal for everyday productivity tasks.

Trial limit:

50 messages per 3 hours

Pro limit:

250 messages per 3 hours

Models:

  • Mistral-small-2503

  • Gemini 2.0 Flash

  • GLM 4.7 Flash

  • Grok-4-fast-non-reasoning

  • Gemma-3-27b

  • Grok-3-mini

  1. Balanced Model

Mid-tier models that balance speed, quality, and cost. Well-suited for writing, analysis, summarization, research assistance, and tasks that require more nuance than Fast models provide.

Trial limit:

30 messages per 3 hours

Pro limit:

100 messages per 3 hours

Models:

  • Minimax M2

  • Minimax M2.1

  • GPT-oss-120b

  • Mistral-medium-2505

  • Gemini 2.5 Flash

  • Mistral-Large 3

  • Gemini 3 Flash

  • Kimi-K2-thinking

  • Qwen – DeepMask

  1. Advanced Model

High-capability models for complex reasoning, deep analysis, coding, strategic work, and tasks where output quality is paramount. These models consume significantly more compute and are priced accordingly.

Trial limit:

10 messages per 3 hours

Pro limit:

50 messages per 3 hours

Models:

  • GLM 4.7

  • Kimi K2.5

  • o3-mini

  • Haiku-4.5

  • Qwen3 – STACKIT

  • DeepSeek-v3-0324

  • GPT-4.1

  • GPT-4o

  1. Premium Model

The most powerful and expensive models available. Reserved for high-stakes tasks requiring the absolute best output — executive briefings, complex legal analysis, advanced coding challenges, and cutting-edge research. Due to their extreme computational demands, rate limits are the most.

Trial limit:

10 messages per 3 hours

Pro limit:

50 messages per 3 hours

Models:

  • GPT-5.2

  • Sonnet-4.5

  • Opus-4.5

What Happens When You Reach a Limit?

What Happens When You Reach a Limit?

If you exceed the allowed number of messages for a model category within the 3-hour window, DeepMask will display a notification informing you that the limit has been reached for that category. Here’s what you can do:

  1. Switch Models:

Immediately use any model from a different category. For example, if you’ve used all your Advanced messages, switch to a Balanced or Fast model to continue working:

  1. Wait for the Window to Reset

Rate limits operate on a rolling 3-hour window. As your oldest messages age past the 3-hour mark, capacity is freed up automatically.

  1. Upgrade your plan

Trial users experiencing frequent limits should consider upgrading to the Professional plan, which offers 3–5× higher limits across all categories.

Important: Rate limits are per category, not per model. Using 10 messages across three different Advanced models still counts as 10 Advanced messages.

Enterprise Plan Customers

Enterprise Plan Customers

Organizations on the DeepMask Enterprise plan benefit from customized rate limits tailored to their specific usage patterns and team size. Enterprise limits are configured during onboarding and can be adjusted at any time in consultation with your dedicated Customer Success Manager.

Enterprise features include:

  1. Custom Rate Limits

Per-user, per-team, or organization-wide — configured to your needs.

  1. Priority Model Access

Guaranteed capacity allocation for mission-critical workloads.

  1. Usage Analytics

Detailed dashboards showing token consumption by user, team, model, and time period.

  1. Dedicated Infrastructure

For sovereign model deployments on STACKIT, dedicated compute ensures no resource contention.

To discuss Enterprise rate limits and custom configurations, contact our sales team at sales@deepmask.io or visit deepmask.io/contact-sales.

Prohibited Activities

Prohibited Activities

The Fair Usage Policy exists alongside DeepMask’s Terms of Service. The following activities are strictly prohibited and may result in account suspension or termination:

  1. Automated Abuse

Using bots, scripts, or automated tools to send messages at scale or circumvent rate limits.

  1. API Scraping

Extracting model outputs programmatically through the chat interface without authorization.

  1. Account Sharing

Sharing login credentials to multiply effective rate limits across multiple individuals.

  1. Circumvention

Any attempt to bypass, manipulate, or exploit the rate-limiting system through technical means.

  1. Resale

Using DeepMask’s output to provide AI services to third parties without a commercial agreement.

DeepMask reserves the right to modify rate limits, adjust model categorizations, or take action against accounts engaged in prohibited activities, at any time and at its sole discretion.

Where Your Data Is Processed

Where Your Data Is Processed

Regardless of which model you use or which category it belongs to, all DeepMask inference is processed within the European Union. Our primary infrastructure partner is STACKIT, the sovereign cloud division of the Schwarz Group, with data centers located in Germany.

Additional EU-region deployments are maintained through partnerships with Microsoft Azure, Amazon Web Services (AWS), and Google Cloud — all configured with strict data residency controls that ensure no data leaves EU jurisdiction.

  1. No data is used for model training

Contractually guaranteed with all upstream AI providers.

  1. End-to-end encryption

AES-256 at rest, TLS 1.3 in transit.

  1. GDPR compliant

Fully aligned with the General Data Protection Regulation.

Updates to This Policy

Updates to This Policy

Model categories, rate limits, and pricing are subject to change as we expand our model offerings, onboard new providers, and respond to platform demand. We will notify users of material changes via email and/or in-app notification at least 14 days before they take effect.

The current version of this policy is always available at deepmask.io/fair-usage-policy.