
- One credit is one AI response on a base model. Premium models cost 2–6 credits each.
- At 3 replies per conversation on a 3-credit model, Standard's 4,000 credits is about 444 conversations, not 4,000.
- Overage runs $40 per 1,000 credits: 33% above the Standard rate and 50% above Pro's.
- Removing Chatbase branding is a $1,188/year add-on, more than a year of Hobby.
The plan looks simple until you launch
Chatbase's pricing page looks simple. Four tiers, a credit number next to each one. The confusion starts after you launch, when a plan rated for 4,000 message credits runs dry at something closer to 1,300 conversations.
Nothing shady is happening. The math just isn't the math most people assume when they pick a plan.
What Chatbase costs in 2026
| Plan | Monthly | Annual | Message credits/mo | Seats |
|---|---|---|---|---|
| Free | $0 | $0 | 50 | 1 |
| Hobby | $32 | $384 | 700 | 2 |
| Standard | $120 | $1,440 | 4,000 | 3 |
| Pro | $400 | $4,800 | 15,000 | 5 |
| Enterprise | Custom | Custom | Custom | Custom |
Annual billing is 20% off. Three add-ons matter for budgeting:
- Extra AI agent: $300 per agent, per year
- Auto-recharge credits: $40 per 1,000 credits
- Remove Chatbase branding: $1,188 per year
That last one deserves a flag. Removing a vendor's logo from your own support widget costs more per year than the entire Hobby plan costs to run. If you're a startup that cares how the widget looks next to your brand, the real entry price isn't $32/month.
The part that catches people: a credit is not a reply
This is the whole article, so here it is plainly: one AI response does not reliably cost one credit. What it costs depends on the model you picked.
Per Chatbase's own documentation, credit consumption per response:
| Credits per response | Models |
|---|---|
| 1 | All other models |
| 2 | GPT-5.2, Gemini 2.5 Pro, Gemini 3.1 Pro, Gemini 3.5 Flash, Gemini 3.6 Flash, GLM 5.2, Mistral Medium 3.5 |
| 3 | Grok 3, Claude 4.5 Sonnet, Claude 4.6 Sonnet, GPT 5.6 Terra |
| 4 | Grok 4, GPT-5.5 |
| 5 | Claude 4.5 Opus, Claude 4.6 Opus |
| 6 | Claude 4.7 Opus, Claude 4.8 Opus |
So the Standard plan is "4,000 credits," but it is only 4,000 replies if you stay on the cheapest tier of models. Pick Claude 4.6 Sonnet (a completely reasonable choice for support quality) and 4,000 credits becomes roughly 1,333 replies.
Same plan. Same $120. One dropdown.
Now multiply by replies per conversation
The second multiplier is that support conversations aren't one message long. A visitor asks something, gets an answer, asks a follow-up, gets a clarification. Three or four AI replies in a single resolved conversation is normal, not heavy usage.
Stack the two multipliers on Standard ($120/mo, 4,000 credits):
| Model tier | Credits/reply | Replies/mo | At 3 replies per convo | Effective cost per conversation |
|---|---|---|---|---|
| Base (1 credit) | 1 | 4,000 | ~1,333 convos | ~$0.09 |
| 2-credit model | 2 | 2,000 | ~666 convos | ~$0.18 |
| 3-credit model | 3 | 1,333 | ~444 convos | ~$0.27 |
| 6-credit model | 6 | 666 | ~222 convos | ~$0.54 |
A $120/month plan handling 222 conversations is a different product than a $120/month plan handling 1,333 conversations.
Both are the Standard plan. The only variable is a dropdown most people set once during setup and never revisit.
What happens when you run out
Credits don't roll over, and when you hit the ceiling mid-month you either upgrade or auto-recharge at $40 per 1,000 credits.
Compare that rate to what you already pay:
- Standard: 4,000 credits for $120 → $30 per 1,000 credits
- Pro: 15,000 credits for $400 → $26.67 per 1,000 credits
- Auto-recharge: $40 per 1,000 credits
Overage credits cost 33% more than Standard-plan credits and 50% more than Pro-plan credits. That's a normal SaaS design (it's meant to push you up a tier), but it means a traffic spike is punished twice: once by the spike, once by the rate.
The failure mode to avoid: a product launch drives traffic, credits drain in week two, and you're buying overage at the worst rate for the rest of the month.
Estimating your real bill in four steps
- 1Monthly conversations. For early-stage products, support contacts tend to run well under 10% of active users per month, though it varies with product complexity and how good your docs are.
- 2AI replies per conversation. Use 3 unless you know better.
- 3Credits per reply. From the table above, and check the model you actually selected, not the default.
- 4Multiply.
conversations × replies × credits = monthly credits
500 conversations on a 3-credit model at 3 replies each is 4,500 credits, over Standard's 4,000, so you're on Pro at $400/month or paying overage.
Run this before you pick a tier, not after.
Where Chatbase's model genuinely makes sense
Credit pricing isn't a scam, and it fits some teams well:
- Low, predictable volume. Under a few hundred conversations a month on a cheap model, Hobby or Standard is inexpensive and the ceiling never bites.
- You want model choice. Chatbase lets you run frontier models. That's a real feature, and the credit multiplier is an honest way to price the cost difference through to you.
- Multi-agent setups, if you accept $300/agent/year.
Where it fits badly: unpredictable traffic, or a team that wants one number on the invoice each month. Credits convert a fixed subscription into a variable one, and the conversion rate is set by a dropdown.
The honest summary
Chatbase's list prices are competitive. The thing to understand before committing is that the headline credit number is a ceiling denominated in a unit you don't control directly. Your model choice and your conversation length both spend it faster than one-credit-per-chat intuition suggests.
Do the four-step estimate with your real numbers. If it lands comfortably inside a tier, Chatbase is priced fine. If it lands near the ceiling, price the overage at $40/1,000 and see whether the answer changes.
Frequently asked questions
- Is one Chatbase message credit equal to one chat?
- No. One credit equals one AI response on a base-tier model. Premium models consume 2–6 credits per response, and a single conversation usually contains several responses.
- Do unused Chatbase credits roll over?
- No. Credits reset each billing period.
- What happens when I run out of credits mid-month?
- You upgrade or buy auto-recharge credits at $40 per 1,000, more expensive per credit than any paid plan's included rate.
- How much does it cost to remove Chatbase branding?
- $1,188 per year, as a separate add-on.
- What's the cheapest realistic Chatbase plan for a real product?
- Hobby ($32/mo, 700 credits) works for genuinely low volume. At 3 replies per conversation on a 2-credit model, that's roughly 115 conversations a month.
Sources
- Chatbase pricing page: plan prices, credit allowances, add-ons (verified 2026-08-11)
- Chatbase FAQ / docs: per-model credit consumption (verified 2026-08-11)
One flat price, no surprise invoice
Helin answers from your own docs, hands off honestly when it doesn't know, and caps your bill at a number you picked. Free tier is permanent: 50 AI replies a month, no trial clock.
