Hi, I am looking at selecting OpenAI model in Vapi, comparing o3-mini vs o4-mini. On Vapi it says GPT o4 Mini Cluster costs $0.04 and GPT o3 Mini Cluster costs $0.01. However on OpenAI o4 Mini is slightly cheaper on Output tokens (as seen on screenshot). Can you please explain why this is the case? https://cdn.discordapp.com/attachments/1458745658742804500/1458745658902052875/image.png?ex=6960c243&is=695f70c3&hm=83ebd37129c78ef356c1c0cac700480e1545ddd79f653584c4011b9ea17bf549&

Hi N3squik, Short answer: Vapi’s UI shows Vapi’s per-minute voice infrastructure fee + the provider model cost estimate. The dropdown price you're seeing is Vapi’s combined displayed rate (not just the raw OpenAI output-token price), and it may use internal mappings or cached provider rates that don't match the raw token-level numbers in OpenAI’s public sheet. What to check next: • Compare actual call log breakdown for a test call using each model (Call Logs show provider model cost vs Vapi fee). This shows the true billed amounts. • If you use your own OpenAI key (or Azure), Vapi only charges the $0.05/min platform fee; provider charges come from your key and will match the provider’s pricing. • If you want, share a call_id or a screenshot of the discrepancy and I’ll inspect the call log and pricing breakdown. Relevant docs: Azure OpenAI model naming/regions (if you’re using Azure) — https://docs.vapi.ai/providers/model/azure-openai

Thank you for providing this information. So I have tested you chat logs and they have roughly the same token counts, and o4-mini is just slightly more expensive. I can see that Platform Cost is separate from the model cost. I would of expected o4-mini to be a lot cheaper, according to LLM calculator (https://yourgpt.ai/tools/openai-and-other-llm-api-pricing-calculator), o4-mini should be around $0.0002 instead of $0.0014707, its a lot more expensive on Vapi for some reason. What are the hidden costs? Referenced chat log ids: - 24ad3c4a-8db5-4c7e-a1f0-4fac360a06ff - 6601500e-64b8-4863-8345-4238b575ded8 https://cdn.discordapp.com/attachments/1458745658742804500/1459486460968177819/image.png?ex=69637430&is=696222b0&hm=f17962a8657da6e22d90e5d422b388ebcf7ceba4a141137f8d30ce616ade2c07& https://cdn.discordapp.com/attachments/1458745658742804500/1459486461266100376/image.png?ex=69637430&is=696222b0&hm=2b62ca7ce6b5bb69afa08df0de178754d83bee054a9854880c682bed3b2b2bec&

Hello, Thank you for sharing the details and the chat log references. I completely understand your concern—seeing the o4-mini pricing on Vapi being significantly higher than expected can be confusing, especially when the token counts are similar and you’re comparing it to other calculators. Based on your observation about o1-mini pricing discrepancies, there are several factors that could explain why you're seeing higher costs than expected from external LLM calculators. Understanding Vapi's Cost Structure Vapi provides detailed cost breakdowns that separate different components: • **Model costs**: Direct LLM provider charges • **Platform costs**: Vapi's service fees • **Other component costs**: STT, TTS, and additional services You can access this breakdown using

Chat.cost

for total cost and

Chat.costs

for the detailed breakdown by component type . Potential Cost Factors Several factors could contribute to higher-than-expected costs: At-Cost vs. Markup Vapi charges "at-cost" for STT, LLM, and TTS provider requests, meaning without profit markup. However, there may be additional platform fees for Vapi's orchestration services. Token Counting Differences External calculators may not account for: • System messages and conversation context • Function calling overhead • Prompt engineering tokens added by Vapi • Conversation state management tokens Model-Specific Pricing o1-mini has unique pricing structures that may differ from standard GPT models, and external calculators might not reflect the most current OpenAI pricing. Investigating Your Specific Costs To understand the exact cost breakdown for your chat logs, you should: 1. Examine the detailed cost breakdown using the

Chat.costs

array for both chat IDs 2. Compare model costs vs. platform costs to see where the discrepancy lies

Okay thank you for explaining. FYI I was comparing o3-min vs o4-mini (not o1-mini). I do understand that there are more things that Vapi cold of been doing, but it would of resulted in more tokens. I believe token counting mechanism is the same for both models, so if there was something extra Vapi doing then we would see completely different token counts. And this is chats, so there are only two prices we will see; LLM and Vapi platform (as we can see in the following breakdown) I have checked the costs and here is the breakdown - ### o3-mini

%{
       "completionTokens" => 223,
       "cost" => 0.0014707,
       "model" => %{"model" => "o3-mini", "provider" => "openai"},
       "promptTokens" => 445,
       "type" => "model"
     },
     %{"cost" => 0.005, "type" => "chat"}

### o4-mini

"costs" => [
     %{
       "completionTokens" => 228,
       "cost" => 0.0014949,
       "model" => %{"model" => "o4-mini", "provider" => "openai"},
       "promptTokens" => 447,
       "type" => "model"
     },
     %{"cost" => 0.005, "type" => "chat"}
   ]

Maybe you have a more detailed view for why o4-mini isnt cheaper than o3-mini? It just feels like there are hidden costs/markup pricing on LLM. Also, if I was to use my own OpenAI keys, will I only pay for Vapi platform changes (which is $0.05 per minute)?

Hi N3squik, I guess there is some confusion on your side if you see the openai's website carefully the 2nd column is the cost of cached input tokens please see the all the columns carefully they cost the same so you are getting the same cost for both the runs, even with different models. I am hereby attaching two screenshots of the open ai website and see if you have any further doubts about the pricing. Please look at the columns carefully. Regards Vapi Support https://cdn.discordapp.com/attachments/1458745658742804500/1461380679299236046/Screenshot_2026-01-15_at_8.53.38_PM.png?ex=696a5851&is=696906d1&hm=0c4d9355bda9f1d65199ffa9b419418c21e1ffff9d964414a701087567636f94& https://cdn.discordapp.com/attachments/1458745658742804500/1461380679530053733/Screenshot_2026-01-15_at_8.53.46_PM.png?ex=696a5851&is=696906d1&hm=5a96eb57d667cd84e1a5675325c73d377332a2fe616b641eee50f6c227162478&

Oh yes thank you for noticing it, so they should be the same price. It is quite misleading to see that on Vapi UI they are being different price (showing o3-mini being 4 times cheaper than o4-mini).

Hi there, I will check that with the team, and get back to you as soon as possible. Regards Vapi Support