Skip to main content
DialNexa bills calls on a per-minute basis, starting from the moment a call connects (is answered) until it ends. The per-minute rate is based on the rounded call duration and the billable components used by the agent pipeline.

​
The Per-Minute Model

You do not receive separate invoices for the LLM tokens used, the seconds of audio synthesized, or the transcription compute. DialNexa handles all provider relationships and presents you with a single per-minute rate for the selected pipeline. What the per-minute rate includes:
ComponentDescription
Telephony (PSTN)Carrier charges for connecting the call. Outbound calls have higher carrier costs than inbound.
Transcription (STT)Speech-to-text processing for the full call duration
LLM inferenceToken costs for all turns in the conversation
TTS synthesisVoice generation for all agent turns
Platform infrastructureDialNexa’s orchestration, storage, and monitoring
What is billed separately (not in the per-minute rate):
  • Phone number rental (monthly, per number)
  • Overage charges for exceeding plan concurrency limits
  • Add-on features such as custom voice cloning (Enterprise)

​
What Affects Your Rate

Your effective per-minute cost varies based on several factors:
The LLM model has the largest impact on per-minute cost. More powerful models (GPT-4o) cost significantly more per token than faster, lighter models (GPT-4o Mini, Groq Llama 4, DeepSeek V3). For most voice agent use cases, GPT-4o Mini provides excellent quality at the lowest cost.Cost relative to GPT-4o Mini (approximate):
  • GPT-4o Mini: 1x (baseline)
  • Groq Llama 4: comparable to GPT-4o Mini (varies)
  • DeepSeek V3: low cost, comparable to or below GPT-4o Mini
  • GPT-4o: ~4-6x
ElevenLabs and Cartesia have slightly different underlying synthesis costs. The difference is modest - typically less than 10% of the total per-minute rate.
Outbound calls cost more than inbound calls because of PSTN termination charges (the carrier fee for connecting a call to a destination number). Inbound calls have lower carrier costs since the originating carrier pays the connection fee.
International calls have higher telephony costs than domestic Indian calls. Some international destinations (specific country codes) have premium carrier rates. The per-minute rate for international calls is higher than for domestic.
Higher-tier plans have lower per-minute base rates. Enterprise accounts with committed monthly volume get the lowest rates through custom pricing.

​
Viewing Exact Pricing

Exact per-minute rates are shown in the dashboard under Settings → Billing → Pricing. The pricing page shows:
  • Your current per-minute rate by call direction (inbound/outbound)
  • Your current per-minute rate by LLM model
  • Phone number rental costs
  • Any plan-level discounts applied to your account
For the public pricing page, visit dialnexa.com/pricing. For custom volume pricing, contact [email protected].
Pricing shown in the dashboard is the most accurate source - it reflects your specific plan tier and any negotiated rates. The public pricing page shows standard rates and may differ from your actual rate.

​
How Costs Are Calculated for a Call

DialNexa first rounds the total connected minutes up to the next full minute, then applies the billable components for the agent pipeline: 
Cascaded agent cost = rounded total minutes × (transcriber cost + LLM cost + TTS synthesis cost + telephony cost)
Speech to Speech agent cost = rounded total minutes × (LLM cost + telephony cost)
Speech to Speech agents use a realtime model that listens and speaks directly, so they do not include separate transcriber or TTS synthesis costs. Here is how a typical cascaded call is costed: 
Call duration: 4 minutes 23 seconds → rounded to 5 minutes
Pipeline rate (transcriber + LLM + TTS synthesis + telephony): ₹X.XX/min
Total: 5 × ₹X.XX = ₹Y.YY
Partial minutes are rounded up to the next full minute for billing. For a batch campaign with 1,000 calls averaging 3 minutes each: 
Total call minutes: 3,000 minutes
Cost: 3,000 × per-minute rate

​
Cost Optimization Tips

Use GPT-4o Mini as your default

GPT-4o Mini is the fastest and most cost-efficient model for voice agents. Only upgrade to GPT-4o or Claude when you have specific quality requirements that GPT-4o Mini cannot meet.

Enable Audio Cache

Pre-generating static phrases with Audio Cache eliminates TTS synthesis costs for those phrases. For agents with common greetings and closings, this can reduce TTS costs by 20-40%.

Keep system prompts concise

Longer system prompts consume more LLM tokens per turn. Trim your prompts to essentials - every turn in the conversation includes the full system prompt in the LLM context.

Set max duration limits

Configure a maximum call duration on your agents. A 15-minute cap prevents runaway calls (e.g., an agent stuck in a loop) from accumulating large charges.

Configure appropriate retry limits

For outbound batch campaigns, setting too many retries increases costs without proportionally increasing connection rates. Start with 2 retries and measure whether additional retries are yielding connections.

Monitor per-agent costs

The usage breakdown in Settings → Billing → Usage shows cost per agent. Identify which agents have high cost-per-call and investigate whether they have unusually long calls or are using expensive models unnecessarily.

​
Free Trial Credits

New accounts receive free trial credits that can be used for real calls. Trial credits cover a defined number of minutes and expire based on the terms shown at signup. Trial calls are subject to the same per-minute rate structure as paid calls - the credits simply absorb the cost.

​
Estimating Costs Before Running a Campaign

Use this formula to estimate campaign cost before launching: 
Estimated cost = (number of contacts × expected answer rate × average call duration in minutes) × per-minute rate

Example:
10,000 contacts × 30% answer rate × 3 min average × ₹X.XX/min
= 9,000 call minutes × ₹X.XX/min
= ₹Y,YYY total estimated cost
The per-minute rate for your specific configuration is shown in Settings → Billing → Pricing.