Quick answer: A credit is not always one response. Compare the credit allowance, selected model and actual credit use before calculating a cost per reply. Boei Starter includes 2,000 AI credits from $19/month with annual billing.
TL;DR: Chatbot tools use message allowances, model-based credits or other usage units. Boei uses AI credits, and the selected model affects how quickly they are used. The tables below do not assume 2,000 Boei credits equal 2,000 replies.
Prices last verified against live vendor pricing pages on 2026-07-19. AI chatbot pricing changes often, especially on credit-based platforms; expect the numbers below to drift within a few months.
How Much Does an AI Chatbot Cost Per Message?
Training data limits tell you how much your AI chatbot can learn. But the other half of the equation, and often the bigger cost driver, is how many responses your chatbot can actually give. Every platform handles this differently, and the differences can mean paying 10x more per customer conversation depending on which AI model you use.
I spent a few hours going through the pricing pages, documentation, and support chatbots of 8 leading AI chatbot platforms to figure out what each response actually costs. For Boei's straightforward pricing, see the AI chatbot pricing page. The platforms I reviewed: Boei, Chatbase, Dante AI, FastBots, SiteGPT, CustomGPT, ChatNode, and Wonderchat.
For each platform, I dug into their credit systems, checked the multipliers for different AI models, calculated cost per response at different tiers. Tested their own chatbots to confirm details that weren't clear from the pricing pages alone. Here's what I found.
Three Pricing Models for Chatbot Responses
AI chatbot platforms use three fundamentally different approaches to limit and charge for chatbot responses.
1. Flat messages (1 response = 1 message, regardless of model)
The simplest approach. Every chatbot response counts as exactly one message, no matter which AI model generates it. You know exactly how many customer conversations you can handle.
Platforms using this model: SiteGPT
2. Credit-based with model multipliers
Each response consumes a variable number of credits depending on the AI model used. A response from a budget model might cost 1 credit. The same response from a premium model could cost 5, 20, or even 330 credits. Your advertised credit count can be misleading because it depends entirely on which model you run.
Platforms using this model: Boei, Chatbase, Dante AI, FastBots
3. Unlimited responses
Some platforms advertise unlimited messages, removing response limits entirely. The trade-off is usually a higher base price or limits elsewhere (like fewer chatbots or less training data).
Platforms using this model: Dante AI (Enterprise only), CustomGPT (within query limits)
The Credit Multiplier Problem
The credit-based model creates a significant transparency issue. When Chatbase advertises "2,000 message credits" on their Hobby plan, most people assume that means 2,000 customer conversations. The reality depends heavily on which AI model powers those responses.
Chatbase Credit Costs Per Response
Chatbase's Hobby plan ($32/mo) includes just 500 message credits, cut sharply from the 2,000 credits they used to offer at this tier. Here is what that translates to in real responses:
| AI Model | Credits Per Response | Real Responses from 500 Credits |
|---|---|---|
| Budget models (GPT-4o Mini, Gemini Flash, budget-tier variants) | 1 credit | 500 responses |
| GPT-5.2 / Gemini 3 Pro / other frontier models | 2 credits | 250 responses |
| Claude Sonnet 4.5 / 4.6 | 3 credits | ~167 responses |
| Grok 4 / GPT-5.5 | 4 credits | 125 responses |
| Claude Opus 4.5 / 4.6 | 5 credits | 100 responses |
| Claude Opus 4.7 | 6 credits | ~83 responses |
The Hobby plan gives you 500 responses on the cheapest models and just 83 responses on Claude Opus. That is a 6x difference in value from the same plan. Chatbase's multipliers themselves are moderate, but the 500-credit floor makes cost-per-response worse than every other platform in this comparison at the entry tier. Extra credits and auto-recharge now cost $40 per 1,000 credits, roughly triple what Chatbase charged in early 2026. If you enable auto-recharge (which most businesses need to avoid their chatbot going silent), those overage costs can double your bill in a busy month.
Dante AI Credit Costs Per Response
Dante AI restructured heavily in 2026. The Starter plan is now $40/mo and includes 2,000 credits (down from $29/3,000 credits earlier in the year). The multiplier system was simplified from a 12-tier range to six tiers, and the extreme 330-credit tier for Claude Opus is gone:
| AI Model | Credits Per Response | Real Responses from 2,000 Credits |
|---|---|---|
| GPT-4o Mini / GPT-4.1 Nano / GPT-5.4 Nano | 1 credit | 2,000 responses |
| GPT-4.1 Mini / Gemini Flash / GPT-5 Mini | 2 credits | 1,000 responses |
| Gemini 3 Flash / Claude Haiku 4.5 / GPT-5.4 Mini / GPT-5.6 Luna | 5 credits | 400 responses |
| GPT-4o / GPT-4.1 / GPT-5 / GPT-5.4 / GPT-5.6 Terra / Gemini 3.1 Pro | 10 credits | 200 responses |
| Claude Sonnet 4.6 | 15 credits | ~133 responses |
| Claude Opus 4.6 / 4.7 / GPT-5.5 / GPT-5.6 Sol | 25 credits | 80 responses |
Dante used to charge 330 credits for a single Claude Opus response, which meant your $29 Starter plan gave you only 9 Opus replies for the entire month. That tier is retired: Opus is now 25 credits, so you get 80 responses on the $40 Starter. Better than before, still expensive. Extra credits cost $20 per 1,000 through the dashboard.
FastBots Credit Costs Per Response
FastBots keeps their multiplier system simple. Their Essential plan is now $33/mo (down from $39) and includes 2,000 messages. In 2026 they added a Claude Sonnet tier at 10 credits, so the multiplier range widened from 5x to 10x:
| AI Model | Credits Per Response | Real Responses from 2,000 Messages |
|---|---|---|
| GPT-4o Mini / GPT-4.1 Mini / GPT-4.1 Nano / GPT-5 Mini / GPT-5 Nano / Claude 3.5 Haiku / Gemini 2.5 Flash / o4 Mini | 1 credit | 2,000 responses |
| GPT-4o / GPT-4.1 / GPT-5 / GPT-5.2 / o3 / Gemini 2.5 Pro | 5 credits | 400 responses |
| Claude 4.5 Sonnet | 10 credits | 200 responses |
FastBots is still more straightforward than Dante AI: three tiers instead of six. Budget models cost 1 credit, most frontier models cost 5, Claude Sonnet costs 10. Your 2,000 messages drop to 400 on GPT-5 or 200 on Claude Sonnet. FastBots also added new tiers in 2026 (a $75/mo plan with 5,000 credits and a $333/mo plan with 30,000 credits) alongside the existing $165 Premium plan (down from $199).
Platforms With Flat Message Pricing
SiteGPT
SiteGPT offers two model choices (GPT-4.1 and GPT-4.1 Mini) and counts messages, not credits. Their Starter plan ($39/mo) includes 4,000 messages per month. GPT-4.1 Mini gives roughly 10x more messages per dollar than GPT-4.1 through SiteGPT's calculator, but there is no complex credit system to navigate. Extra messages cost $39 per 5,000 messages.
Boei
Boei Starter includes 2,000 AI credits for $19/month with annual billing ($25 billed monthly). Credit use depends on the selected model. Growth includes 7,000 credits and Business includes 20,000. Measure credit consumption with your own customer questions before deciding which plan covers your expected number of replies.
Normalized Cost Per Response Comparison
This is where it gets interesting. The table below shows the actual cost per response on each platform's entry-level paid plan, calculated for both a budget model and a premium model.
Using a Budget Model (GPT-4o Mini or equivalent)
| Platform | Plan Price | Included Credits/Messages | Responses (Budget Model) | Cost Per Response |
|---|---|---|---|---|
| Boei | $19/mo annual | 2,000 AI credits | Model-dependent | Model-dependent |
| SiteGPT | $39/mo | 4,000 messages | 4,000 (GPT-4.1 Mini) | $0.0098 |
| FastBots | $33/mo | 2,000 messages | 2,000 (Mini models at 1x) | $0.0165 |
| Dante AI | $40/mo | 2,000 credits | 2,000 (GPT-4o Mini at 1x) | $0.0200 |
| Chatbase | $32/mo | 500 credits | 500 (budget models at 1x) | $0.0640 |
Using a Mid-Tier Model (GPT-4.1 Mini or 4o Mini equivalent, 1-2 credit tier)
| Platform | Plan Price | Responses (Mid-Tier Model) | Cost Per Response |
|---|---|---|---|
| Boei | $19/mo annual | Model-dependent | Model-dependent |
| SiteGPT | $39/mo | 4,000 (GPT-4.1 Mini) | $0.0098 |
| FastBots | $33/mo | 2,000 (Mini at 1x) | $0.0165 |
| Dante AI | $40/mo | 1,000 (GPT-4.1 Mini at 2x) | $0.0400 |
| Chatbase | $32/mo | 500 (at 1x) | $0.0640 |
Using a Premium Model (GPT-5 or GPT-4o tier)
| Platform | Plan Price | Responses (Premium Model) | Cost Per Response |
|---|---|---|---|
| Boei | $19/mo annual | Model-dependent | Model-dependent |
| SiteGPT | $39/mo | 4,000 (GPT-4.1) | $0.0098 |
| FastBots | $33/mo | 400 (GPT-5 at 5x) | $0.0825 |
| Chatbase | $32/mo | 250 (GPT-5 at 2x) | $0.1280 |
| Dante AI | $40/mo | 200 (GPT-5 at 10x) | $0.2000 |
The Extreme Case: Top-Tier Models (Claude Opus)
| Platform | Model | Responses on Entry Plan | Cost Per Response |
|---|---|---|---|
| Boei ($19/mo annual) | Selected model | Model-dependent | Model-dependent |
| Chatbase ($32/mo) | Claude Opus 4.5/4.6 (5 credits) | 100 responses | $0.32 |
| Chatbase ($32/mo) | Claude Opus 4.7 (6 credits) | ~83 responses | $0.39 |
| Dante AI ($40/mo) | Claude Opus 4.6/4.7 (25 credits) | 80 responses | $0.50 |
| FastBots ($33/mo) | Claude Sonnet 4.5 (10 credits) | 200 responses | $0.165 |
The Hidden Cost: Overage Pricing
Running out of credits mid-month is where costs can really spike.
Chatbase now charges $40 per 1,000 extra credits, up sharply from the $12-14 range earlier in 2026. With auto-recharge enabled (which most businesses need to avoid their chatbot going silent), this adds up fast. On Claude Opus, 1,000 auto-recharge credits at $40 gives you around 170 responses, which works out to $0.24 per response on top of whatever you are already paying for your plan.
FastBots sells extra message add-ons through their pricing page. Predictable per-block cost, but the total climbs quickly at premium-model multipliers.
Dante AI charges $20 per 1,000 credits through the dashboard for extra credit top-ups. Predictable, but on Claude Opus at 25 credits per response, $20 buys you 40 extra Opus replies.
Boei: Check the credit balance and the upgrade or extra-credit options in your account. Adding a chatbot is not a promise of a fixed number of extra replies.
SiteGPT charges $39 per extra 5,000 messages, which is a clean and predictable overage model.
What "Unlimited" Actually Means
Some platforms advertise unlimited messages, but this deserves scrutiny.
Dante AI only offers truly unlimited responses on their Enterprise plan (custom pricing, typically $1,000+/mo). Their standard plans (Starter $40, Advanced $120, Pro $400) all have credit limits.
CustomGPT includes query allowances within their plans, but reaching higher limits requires jumping from $99/mo to $499/mo.
In practice, no platform at the $19-100/mo range offers genuinely unlimited AI responses. They all have limits, just expressed differently.
Practical Scenarios
Small Business: ~500 customer conversations per month
A small business website with moderate traffic, getting about 500 chatbot conversations monthly. Average conversation length: 3-4 exchanges (so roughly 500 responses per month if counting single responses, or 1,500-2,000 if counting full conversation threads).
At 500 responses/month using a mid-tier model:
- Boei: Estimate credits using the selected model and this workload. The number of replies alone does not determine the right plan.
- SiteGPT ($39/mo): Well within 4,000 limit. Cost: $39/mo
- FastBots ($33/mo): Within 2,000 limit at 1x. Cost: $33/mo
- Dante AI ($40/mo): 500 x 2 = 1,000 credits at Mini rates. Within 2,000 limit. Cost: $40/mo
- Chatbase ($32/mo): 500 responses at 1 credit each hits the Hobby plan's 500-credit ceiling exactly. Any spike triggers auto-recharge at $40/1,000 credits. Cost: $32/mo baseline, unpredictable in practice.
Growing Business: ~2,000 conversations per month
A growing e-commerce store or SaaS product with 2,000 chatbot responses monthly using a premium model (GPT-5 tier):
- Boei: Estimate credits using the selected model and this workload. The number of replies alone does not determine the right plan.
- SiteGPT ($39/mo): Within 4,000 limit. Cost: $39/mo
- Dante AI ($40/mo): 2,000 x 10 credits = 20,000 credits needed. Starter has 2,000. Would need Pro plan at $400/mo (30,000 credits) or Advanced at $120/mo (8,000, still not enough)
- FastBots ($33/mo): 2,000 x 5 = 10,000 credits needed. Essential has 2,000. Would need Premium plan at $165/mo (10,000 credits)
- Chatbase ($32/mo): 2,000 x 2 = 4,000 credits needed. Hobby has 500. Would need Standard plan at $120/mo (4,000 credits)
High-Volume: ~5,000 conversations per month
A busy support operation handling 5,000 responses monthly on a premium model:
- Boei: Estimate credits using the selected model and this workload. The number of replies alone does not determine the right plan.
- SiteGPT: Growth plan at $79/mo for 10,000 messages
- FastBots: 5,000 x 5 = 25,000 credits. Premium at $165/mo covers 10,000; needs the higher $333/mo tier (30,000 credits) or extras
- Chatbase: 5,000 x 2 = 10,000 credits. Would need Standard ($120/mo, 4,000) plus significant auto-recharge, or Pro ($400/mo, 15,000)
- Dante AI: 5,000 x 10 = 50,000 credits. Would need Pro plan at $400/mo (30,000) plus $400 in overage credits
The Model Quality Trade-Off
Credit-based platforms create a real dilemma: you can use a cheap model to maximize your response count, but your customers get lower-quality answers. Or you can use a premium model for better answers but burn through your credits in days.
Key Findings
-
Credit-based pricing is inherently unpredictable, and the disparities widened in 2026. All three credit-based platforms in this comparison (Chatbase, Dante AI, FastBots) raised Claude Sonnet multipliers between 2x and 3x during 2026. The same plan can give you anywhere from 80 to 2,000 responses depending on model choice.
-
Boei Starter includes 2,000 AI credits, not a guaranteed 2,000 replies. Model choice affects credit consumption.
-
Chatbase cut its Hobby credit allowance from 2,000 to 500 in 2026 and tripled auto-recharge to $40 per 1,000 credits. The Hobby plan is no longer competitive on cost-per-response: 500 credits at 1 per response means the plan runs dry inside 500 replies, and the overage burn is severe.
-
Dante AI retired its 330-credit Opus tier, but its overall pricing moved up. Starter went from $29 to $40, credit allowance dropped from 3,000 to 2,000, and premium-model multipliers roughly doubled. Better on Opus, worse everywhere else.
-
FastBots kept the cleanest credit system and dropped Essential from $39 to $33, but added a 10-credit tier for Claude Sonnet that puts Sonnet at $0.165 per response on the entry plan.
-
SiteGPT offers strong value with simple pricing at $39 Starter and $79 Growth. Prices held steady through 2026.
-
Check how extra usage is handled. Compare the cost of extra credits or a higher plan, and check whether automatic charges are enabled.
-
Always calculate cost per actual response, not cost per credit. A plan with "8,000 credits" could mean 8,000 responses or 320 responses depending on the model. The credit number on the pricing page is almost meaningless without knowing which model you will use.
How to Compare Platforms Yourself
When evaluating any AI chatbot platform's message pricing, ask these questions:
- How many credits or messages does the plan include?
- Which AI model will I actually use in production?
- How many credits does that specific model consume per response?
- What is the real cost per response (plan price / actual responses)?
- What happens when I run out? What do overages cost?
- Can I switch models without changing my budget?
Run these numbers with the model you plan to use, not the cheapest model the platform offers. That gives you the true cost comparison. For the other side of the equation, see our companion article on AI chatbot training data limits compared. And if you want to understand the broader value of adding a chatbot, read about AI chatbot benefits.
FAQ
What is a message credit in AI chatbot pricing?
A message credit is a unit that platforms use to measure chatbot responses. On some platforms, one credit equals one response. On others, a single response can consume 1 to 330 credits depending on which AI model generates it. Always check the credit multiplier for the model you plan to use.
Why does the AI model affect the price per response?
Different AI models have different costs to run. Boei credit use depends on the selected model too. Test the models you plan to use before estimating your monthly allowance.
How many AI chatbot responses does a small business need per month?
Most small business websites need 300-1,000 responses per month. A mid-size e-commerce store or SaaS product might need 2,000-5,000. Start with an entry-level plan and track your actual usage for the first month before upgrading.
What happens when I run out of message credits?
It depends on the platform. Check whether the chatbot stops at its allowance, whether extra usage is charged automatically, and whether you can buy extra credits or upgrade. In Boei, review the credit options in your account before estimating extra replies.