Comparison
An AI interview assistant without a subscription
Most AI interview assistants charge monthly. A bring-your-own-key app flips that: you pay once for the app and pay an AI provider directly for what you use. This page explains how that works, how to get an API key, and what an AI interview assistant with a one-time payment actually costs per answer and per month.
Disclosure: we make Interview LLM, a $39 one-time, bring-your-own-key app. API prices below come from OpenAI's, Anthropic's and Google's pricing pages, and subscription prices from each vendor's site, all checked on Nov 14, 2026. Product names are trademarks of their owners, and this page isn't affiliated with them.
How bring-your-own-key apps work
"Bring your own key" (BYOK) means the app doesn't resell AI access. You create an account with a model provider, generate an API key, and paste it into the app. When you send a screenshot, the app calls the provider with your key, and the provider bills you for the tokens used. The app maker never sees that traffic or charges you per request.
In Interview LLM, you choose an OpenAI, Anthropic (Claude) or Google (Gemini) model, and you can pick a different model for each stage: extracting the problem from the screenshot, writing the solution, and debugging. The key stays on your machine. The license is $39 once (shown against $69), with no subscription.
Getting an API key
The steps are similar at each provider. Interfaces change, so follow the provider's current docs if a menu has moved.
OpenAI
Sign in to the OpenAI API platform (this is separate from your ChatGPT account settings), add a payment method or buy prepaid credits in the billing section, then create a secret key on the API keys page. Copy it right away; it's shown once.
Anthropic (Claude)
Sign up for the Claude Console, buy credits under billing, and create a key under API keys. A Claude Pro or Max chat subscription doesn't include API credit.
Google (Gemini)
Open Google AI Studio and use "Get API key" to create a key tied to a Google Cloud project. Several Gemini models have a free tier with rate limits; to use paid models or higher limits, turn on billing for that project.
Whichever you use, set a monthly spending limit or budget alert if the provider offers one, and revoke the key if it's ever exposed.
What one answer costs
Providers charge per million tokens, with output priced higher than input. Our working assumption for one answer, combining the extraction and solution calls:
- Input: about 3,000 tokens, including the screenshot and instructions.
- Output: about 2,500 tokens: the extracted problem, the code, an explanation and complexity analysis.
Worked example with Claude Sonnet 5.5 at $2 input and $10 output per million tokens: 3,000 × $2 / 1,000,000 = $0.006, plus 2,500 × $10 / 1,000,000 = $0.025, for $0.031 per answer, about three cents. The same formula across models, at list prices as of Nov 14, 2026:
| Model | Input / output per 1M tokens | Per answer | 60 answers a month | 300 answers a month |
|---|---|---|---|---|
| GPT-6 Luna | $0.10 / $0.50 | $0.0016 | $0.09 | $0.47 |
| GPT-5.4 mini | $0.75 / $4.50 | $0.0135 | $0.81 | $4.05 |
| GPT-5.4 | $2.50 / $15 | $0.045 | $2.70 | $13.50 |
| GPT-6 Astra | $10 / $50 | $0.155 | $9.30 | $46.50 |
| Claude Haiku 5.5 | $0.10 / $0.50 | $0.0016 | $0.09 | $0.47 |
| Claude Sonnet 5.5 | $2 / $10 | $0.031 | $1.86 | $9.30 |
| Claude Opus 5.5 | $4 / $20 | $0.062 | $3.72 | $18.60 |
| Gemini 3 Flash Preview | $0.50 / $3 | $0.009 | $0.54 | $2.70 |
| Gemini 3.1 Pro Preview | $2 / $12 | $0.036 | $2.16 | $10.80 |
| Gemini 4 Carbon (through Nov 30) | $2 / $10 | $0.031 | $1.86 | $9.30 |
| Gemini 4 Carbon (from Dec 1, 2026) | $4 / $20 | $0.062 | $3.72 | $18.60 |
Sixty answers a month is about two problems a day; 300 is about ten a day, a heavy prep month. The table is a price reference across each provider's current lineup; the models you can pick inside Interview LLM are listed in its pricing section.
What pushes the cost up
- Images. Screenshots are billed as input tokens, and larger or higher-resolution images cost more. Two screenshots per problem roughly doubles the image part.
- Reasoning tokens. Models that "think" before answering bill those hidden reasoning tokens as output. If an answer generates 10,000 output tokens instead of 2,500, Claude Sonnet 5.5 goes from $0.031 to $0.106 per answer ($0.006 + $0.100), and 300 answers a month from $9.30 to $31.80. GPT-6 Astra goes from $0.155 to $0.53 per answer, or $159 for 300.
- Follow-ups. Each debugging screenshot is another request.
Choosing a model per stage keeps this down: a cheap model is often enough to read the problem from a screenshot, and you can save a stronger model for the solution.
Subscription vs one-time: the break-even
A one-time license wins once the monthly fees you avoid exceed the license price. The formula:
Break-even months = $39 / (subscription price per month − your monthly API cost)
Using $9.30 a month in API costs (300 answers on Claude Sonnet 5.5) and subscription prices as of Nov 14, 2026:
- Against $299/month (Interview Coder Monthly Pro): $39 / $289.70 = 0.13 months. Cheaper from the first month.
- Against $59/month (Interview Solver monthly): $39 / $49.70 = 0.78 months. Also cheaper in the first month.
- Against $25/month (Final Round AI's "$25+/mo" starting price): $39 / $15.70 = 2.5 months. Cheaper from month 3.
The formula also shows when BYOK loses: if your API bill is higher than the subscription, the one-time license never catches up. Running GPT-6 Astra with long reasoning at $159 a month costs more than a $59 subscription, although the two aren't strictly comparable because subscription vendors don't name the models they run. For a full side-by-side of plans, see AI coding interview assistants compared, or the Interview Coder pricing breakdown for a 12-month view.
The trade-offs of bring your own key
- Setup. You create a provider account, add billing and paste a key. It's a few minutes, but it's more than signing into an app.
- A billing account. Paid models need a card or prepaid credits with the provider, on top of the license.
- Separate from chat subscriptions. ChatGPT Plus, Claude Pro and similar plans don't cover API usage.
- Variable cost. Your bill depends on the model and how much you use it. Spend limits help.
- Rate limits. New API accounts may start with lower limits, which rarely matters for one person solving problems.
- The upside. You choose and switch models, see exactly what each request costs, and your screenshots go straight to the provider rather than through the app maker's servers. With Interview LLM, the included source code shows exactly what's sent.
Using it the right way
Many employers and assessment platforms prohibit outside help during live interviews and online assessments, so follow the rules you agreed to. Where a BYOK app pays off most is practice: screenshot a problem, study the worked solution and its complexity analysis, then solve it again yourself. If that fits how you prepare, you can see Interview LLM's pricing.
Sources (checked Nov 14, 2026)
- OpenAI API pricing: developers.openai.com/api/docs/pricing
- Anthropic API pricing: platform.claude.com/docs/en/about-claude/pricing
- Google Gemini API pricing: aistudio.google.com/docs/pricing
- Interview Coder plans: interviewcoder.co
- Interview Solver pricing: interviewsolver.com/pricing
- Final Round AI: finalroundai.com
FAQ
Does my ChatGPT Plus or Claude Pro subscription cover API usage?
No. Consumer chat subscriptions and API usage are billed separately. You need an API account with its own payment method or prepaid credits.
How much does a bring-your-own-key interview app cost per month?
With our assumptions (about 3,000 input and 2,500 output tokens per answer), 300 answers a month costs about $2.70 on Gemini 3 Flash Preview, $9.30 on Claude Sonnet 5.5 and $46.50 on GPT-6 Astra at list prices as of Nov 14, 2026. Long reasoning outputs can multiply that.
Can I use a free API tier?
Google AI Studio offers a free tier for several Gemini models, with rate limits. Read its terms first, including how prompts may be used on the free tier.
Is my API key safe in Interview LLM?
The key and your session data stay on your machine, and the app only calls the provider you chose. The full source code is included, so you can check exactly what it sends. You can revoke a key in your provider's dashboard at any time.