Credits
Every credit pays for two independent AI classifications per response — genuine dual-rater reliability, not a single-model shortcut. The transparent formula lets you predict costs exactly before each run.
How Credits Work
Credits are consumed each time you run a coding job. The number of credits required depends on the number of responses you want to code, the quality tier you choose, and a small overhead based on how much training data you include.
Credits are deducted at the start of each run. If a run is cancelled or fails partway through, unused credits are automatically refunded based on the number of responses that were not processed.
Credit Formula
The credit cost for a coding run is calculated as:
training_overhead = ceil(min(training_tokens, 100,000) / 10,000)
base_cost = responses + training_overhead
credits = ceil(base_cost x tier_factor)| Component | What It Represents | Range |
|---|---|---|
responses | Number of responses to code | 1 - 50,000 |
tier_factor | User-facing quality tier multiplier | 1.0x - 3.0x |
training_overhead | Small fixed cost based on training-data size | 0, 3, 5, or 10 credits |
Model Tiers
Choose the quality tier that matches your research needs. Higher tiers use more capable AI models and consume more credits per response.
| Tier | Factor | Best For |
|---|---|---|
| Budget | 1.0x | Simple categories, exploratory coding, high volume |
| Standard | 1.5x | Default tier, good balance of cost and quality |
| Quality | 3.0x | Complex categories, nuanced responses, publication-ready |
Start with Standard. For most research, Standard tier provides excellent results. Only upgrade to Quality if you're seeing lower-than-expected agreement rates or have particularly nuanced categories.
Training Data Overhead
Training examples and category descriptions add a small fixed overhead to each run. This overhead is additive, not multiplicative, so large studies are not punished for having a well-developed coding guide.
| Training Data Size | Added Cost | Description |
|---|---|---|
| None | +0 credits | No training examples |
| Small | +3 credits | About 25,000 tokens |
| Medium | +5 credits | About 50,000 tokens |
| Large | +10 credits | About 100,000 tokens (max counted) |
What's a token? Roughly 4 characters or 0.75 words in English. A typical training example of 50 words is about 65 tokens. The qualcode.ai interface shows your training data size before you start a run.
Example Calculations
Here are some common scenarios to help you estimate costs:
Small Study, Standard Settings
training_overhead = 0
base_cost = 500
credits = ceil(500 x 1.5) = 750Large Study, Budget Model, No Training Data
training_overhead = 0
base_cost = 5,000
credits = ceil(5,000 x 1.0) = 5,000Medium Study, Quality Model, Extensive Training
training_overhead = 5
base_cost = 1,005
credits = ceil(1,005 x 3.0) = 3,015Small Pilot With Training Examples
training_overhead = 3
base_cost = 203
credits = ceil(203 x 1.5) = 305See exact costs before running: The coding run dialog shows the exact credit cost based on your settings before you confirm. No surprises.
Free Credits
Every new qualcode.ai account receives up to 500 free credits:
- 50 credits immediately upon registration
- +450 credits after verifying your email address
This is enough to code approximately 330 responses at Standard quality, or 500 responses at Budget quality - plenty to test the platform with your own data before purchasing.
No credit card required. Sign up and start coding immediately with your free credits. Verify your email to unlock the full 500 credits.
Getting More Credits
When you need more credits than the free tier provides, you can either buy an individual Researcher package or contact us for an institutional annual license.
See Pricing or email sales@qualcode.ai.
Purchased credits never expire. Free credits expire 12 months after issuance if unused. Individual purchases are pay-as-you-go, with no subscription or monthly fee.
Auto-Suggest Credits
The Auto-Suggest Coding Guide feature has separate pricing from coding runs. Auto-suggest always uses our highest-capability suggestion models (GPT-5.2 and Claude Opus 4.5), so there is no model tier selector for this workflow.
| Mode | Base Cost (300 samples) | Description |
|---|---|---|
| Quick | 29 credits | Direct dual-rater analysis with high reasoning effort |
| Thorough | 58 credits | Deeper multi-step analysis with a 2x mode multiplier |
Cost is calculated as max(5, floor((5 + 0.08 × sample_size) × mode_factor)), where Quick uses 1.0x and Thorough uses 2.0x.
No model tier selection for Auto-Suggest. Unlike coding runs, Auto-Suggest always uses our most capable models to ensure high-quality category suggestions.
Credit Refunds
Credits are refunded automatically in these situations:
- Cancelled runs: Unprocessed responses are refunded immediately
- Failed runs: If a system error prevents completion, unprocessed responses are refunded
- Rejected responses: Pre-filtered responses (empty, spam) are not charged
Successfully processed responses are not refunded. Once a response has been coded by both AI raters, those credits have been consumed. Review your settings carefully before starting large runs.
Tracking Your Usage
Monitor your credit balance and history:
- Dashboard header: Current balance always visible
- Transaction history: Complete record of credits earned, spent, and refunded
- Per-run costs: Each coding run shows its credit cost
Related: Learn about the Dual-Rater Methodology that powers qualcode.ai.