Eval directory · Voice AI

Evals for Telnyx

Eval coverage for Telnyx, mapped from its public product surface.

About Telnyx

Telnyx is a communications platform that sells voice, messaging, and AI infrastructure through APIs and a self-service Mission Control portal. It markets an owned, vertically integrated stack spanning its private network, programmable identity/compliance control plane, and colocated compute for real-time voice AI agents. Products are priced pay-as-you-go or via volume-based contracts, covering phone numbers, SIP trunking, Voice API, messaging, WhatsApp, video, IoT, and inference.

Industry

communications platform (CPaaS) with voice AI agent infrastructure

Website

telnyx.com

Use the eval library for Telnyx

We'll build out the full library — runnable test cases with inputs, expected behavior, and pass/fail checks — in your Corsac workspace.

Related in Voice AI

All evals →

More Voice AI eval libraries

Coverage map

What would you measure for Telnyx?

6 scoring areas · 24 capabilities mapped · grounded in 8 cited pages

Every eval set is graded on

Adversarial robustness
Workflow quality
Safety gates
Operator quality

Pass/Fail + LLM judge 1–5 · critical severity flags · negative controls

Programmable Voice API

Call control over the Voice API and TeXML, including inbound and outbound per-minute charges that stack on top of SIP Trunking fees, plus optional in-call features.

“Inbound and outbound WhatsApp calls to 2 billion users through the same Telnyx number” telnyx.com

Mapped capabilities

4 capabilities

Outbound and inbound call pricing composition
$0.002/min Voice API charge added to the applicable SIP Trunking fee; identical $0.002 treatment for TeXML; rates vary by direction, number type, and destination.
Optional in-call features
Call recording, recording storage, conference per-participant billing, SIP interface, call transfer per invocation, browser/app calling, secure media.
Answering machine and deepfake detection
Standard vs premium answering machine detection tiers and per-invocation deepfake detection on live calls.
Number-type and destination sensitivity
Local vs toll-free inbound handling and the global SIP Trunking price sheet as the source for destination-specific rates.

Illustrative example

Input: What is my all-in per-minute cost to place an outbound call to a US local number with the Voice API? We are on pay-as-you-go.
Expected behavior: States the $0.002/min Voice API charge is additive to the separate SIP Trunking outbound fee, not the total, and points to the global SIP Trunking price sheet for the destination rate rather than quoting one flat number.

Elastic SIP Trunking

Carrier-grade trunking on the owned network, sold either per minute or via inbound concurrency channels, with capacity and regulatory add-ons.

“Telnyx owns the full stack from carrier network to AI inference” telnyx.com

Mapped capabilities

4 capabilities

Per-minute vs inbound channel models
Metered local and toll-free inbound rates against banded channel pricing for unlimited concurrent inbound calls.
Outbound rates and CPS thresholds
International calls starting at $0.005/min, free outbound toll-free, and call-per-second thresholds that govern outbound capacity.
Included vs paid trunk features
Free call concurrency, secure trunking, and T.38 fax support versus paid call recording and per-number emergency calling.
Owned-network positioning
Bare-metal fiber, real-time transport fabric, and media/edge processing as the stated basis for avoiding the public internet.

Messaging and WhatsApp

SMS, MMS, RCS, and WhatsApp messaging and calling, billed per message part or per minute with carrier and Meta fees layered on top.

Mapped capabilities

4 capabilities

Per-message-part billing
Outbound and inbound SMS/MMS rates applied per part, plus separately stated carrier fees; character limits determine part counts.
Sender type selection
Local including 10DLC, toll-free, short code, and RCS senders, and how destination and carrier affect the applied rate.
Volume discount bands
Automatic discounts above 100M messages/month with distinct outbound and inbound ladders across the next 150M, 250M, 500M, and 1B parts.
WhatsApp Business and Calling
Platform usage per WhatsApp message and per WhatsApp call minute on the same Telnyx number and connection, with separate Meta messaging and business-initiated calling fees.

Illustrative example

Input: We will send 300 million outbound SMS parts next month from 10DLC numbers. What do we pay per part, and is anything billed on top?
Expected behavior: Explains billing is per message part across the published bands rather than one blended rate: the first 100M at the $0.004 pay-as-you-go rate, the next 150M at $0.0020, the remaining 50M at $0.0015, with carrier fees charged separately.

Numbers, Identity, and Compliance

Number inventory and the programmable identity/compliance control plane, covering acquisition terms, reputation, and verification surfaces.

Mapped capabilities

4 capabilities

Number types and terms
Local and toll-free from $1/month, 800 TFN with $500 OTC covering a 12-month commitment then $40 MRC, short code and vanity short code monthly rates.
Volume number discounts and capability add-ons
Automatic discounts past 50 numbers/month with bands down to $0.25, and the $0.10/month SMS/MMS capability surcharge.
Reputation monitoring and remediation
Enterprise reputation MRC, per-number per-check pricing, and per-number remediation.
Identity APIs
Number Lookup API, Verify API, and Branded Calling as the listed identity products.

Voice AI Agent Stack

Real-time agent infrastructure with GPUs colocated on the live communications path, exposed through conversational AI, speech services, and inference.

“We bridge the gap by colocating GPUs with the live communications path.” telnyx.com

Mapped capabilities

4 capabilities

Speech-to-text vendor selection
Per-minute STT options spanning Telnyx, Deepgram Nova 2/Nova 3/Flux, Google, Azure, and AssemblyAI at materially different rates.
Text-to-speech selection
Per-character TTS pricing across Telnyx and standard (Amazon Polly) voices.
Conversational AI and inference
Conversational AI as a Voice API optional feature with its own comprehensive pricing, alongside the standalone Inference product.
Colocated real-time compute
GPUs colocated with the live call path, AI-native orchestration, storage, and observability as the stated agent runtime layers.

Pricing Model and Account Operations

How the commercial model and self-service portal work: choosing pay-as-you-go versus contract, earning automatic discounts, and getting support.

Mapped capabilities

4 capabilities

Pay-as-you-go vs volume-based contracts
Usage-only billing with automatic scale discounts against committed contracts with negotiated discounted rates and predictable monthly payments.
Mission Control portal
Self-service setup, management, monitoring, and account top-up.
Automatic discount thresholds
Discounts applied automatically as calling, messaging, and number volumes cross published thresholds, without renegotiation.
Support entitlements by tier
Free 24/7/365 in-house chat and phone support for all, versus dedicated CSM, prioritized ticketing, and NOC support on contract.

Coverage is mapped from Telnyx's public pages (8 crawled). Examples are illustrative, not real test cases. The runnable eval library — graded inputs, expected behavior, and pass/fail checks — is built when you request it above.

Frequently asked questions

What do the Corsac evals for Telnyx test?+

The coverage map is generated from Telnyx's own public product surface (communications platform (CPaaS) with voice AI agent infrastructure): 6 scoring areas — Programmable Voice API, Elastic SIP Trunking, and Messaging and WhatsApp, and more — spanning 24 mapped capabilities, each graded on adversarial robustness, workflow quality, safety gates, and operator quality once the library is built.

How are the Telnyx evals scored?+

Every case generated for Telnyx — across Programmable Voice API and Elastic SIP Trunking and the other mapped areas — is graded with pass/fail checks plus an LLM judge scoring 1–5 against its expected behavior, with critical-severity flags and negative controls. Only judge-passed evals are published.

How many test cases does the Telnyx library include?+

The full Telnyx library is built on request. The coverage map spans 6 areas and 24 capabilities (for example, Outbound and inbound call pricing composition and Optional in-call features under Programmable Voice API); each becomes graded test cases — inputs, expected behavior, pass/fail checks — in your Corsac workspace.

How do I run these evals against Telnyx or my own agent?+

Request the library with your work email above. We'll build out all 6 mapped Telnyx areas and set them up in a Corsac workspace, where you can run every test case against Telnyx or your own agent with your own data.