Young fern fronds and moss with dew, seen through fluted glass

Energy and environment

Phone agents designed to use less energy.

Phone agents designed to use less energy.

Phone agents designed to use less energy.

Most of every call runs on small AI models we train ourselves. A large model steps in only for the open-ended parts. Here is where the electricity goes, what the research says, and what we are measuring next.

Most of every call runs on small AI models we train ourselves. A large model steps in only for the open-ended parts. Here is where the electricity goes, what the research says, and what we are measuring next.

Most of every call runs on small AI models we train ourselves. A large model steps in only for the open-ended parts. Here is where the electricity goes, what the research says, and what we are measuring next.

Contact Sales

An engineering description, not a pledge.

An engineering description, not a pledge.

An engineering description, not a pledge.

Most of a phone call is handled by small models we train and host ourselves. The open-ended turns run on a large language model. Where a number on this page is someone else’s, we name them. Where it is ours, we will publish how we measured it.

Most of a phone call is handled by small models we train and host ourselves. The open-ended turns run on a large language model. Where a number on this page is someone else’s, we name them. Where it is ours, we will publish how we measured it.

Most of a phone call is handled by small models we train and host ourselves. The open-ended turns run on a large language model. Where a number on this page is someone else’s, we name them. Where it is ours, we will publish how we measured it.

Self-hosted speech recognition

Small models on our GPUs

Scripted turns skip the LLM

Cached speech

Reviewed September 2026

Where a call’s compute goes

Four jobs on every turn. Very different footprints.

Four jobs on every turn. Very different footprints.

Four jobs on every turn. Very different footprints.

A voice agent listens, decides, writes and speaks, dozens of times per call. Model size sets the energy of each job, and only one of the four needs a large model.

A voice agent listens, decides, writes and speaks, dozens of times per call. Model size sets the energy of each job, and only one of the four needs a large model.

LISTEN

0.6–1.7B

0.6–1.7B

0.6–1.7B

parameters, speech recognition

Live transcription runs on speech recognition models we host ourselves, alongside a self-hosted commercial engine.

Runs on our GPUs

DECIDE

<1B

<1B

<1B

parameters, what happens next

The conversation flow and small models we trained decide the next step: intent matching, yes-or-no and option detection. Scripted lines, including the greeting and the opening, are spoken verbatim with no language model call.

Runs on our GPUs

WRITE

120B

120B

120B

parameters, open-ended turns

Turns that need free-form language are written by a large open-weights model, with frontier models available per agent. Part of this inference runs at third-party providers. This is the heavy job.

Large language model

SPEAK

Cached

Cached

Cached

text-to-speech, third-party

Speech is synthesized by a text-to-speech provider. Scripted lines are pre-generated when a call starts, and identical utterances are served from a voice cache.

Third-party provider

Also on our GPUs

Also on our GPUs

Also on our GPUs

Entailment checks, embeddings, query rewriting, dialogue segmentation and voice activity detection. Small models we host ourselves, most under one billion parameters.

Entailment checks, embeddings, query rewriting, dialogue segmentation and voice activity detection. Small models we host ourselves, most under one billion parameters.

Calls are short

Calls are short

Calls are short

51% of cold calls end inside the first minute

In Klenty’s study of 25,000 cold calls, 12,845 ended within 60 seconds. Gong puts a successful cold call at about 5 minutes 50 seconds and an unsuccessful one at roughly half that. The start of a call, the greeting, the identification and the opening, is exactly the part that runs on scripted turns and small models. On a typical sales call, a large share of the turns never reaches the large model.

Klenty, How long should a cold call last, 25,000-call study. Gong, cold call statistics. Industry data, not ours. We will publish our own share of turns, see below.

Design choices

How we spend less of it.

How we spend less of it.

How we spend less of it.

A single moss-covered river stone on a cream surface - small, quiet, and enough

01

Scripted turns skip the language model

Scripted turns skip the language model

Scripted turns skip the language model

The greeting, the opening and every scripted line are spoken verbatim from the conversation flow. The language model is called only for open-ended turns.

02

Small models around the conversation

Small models around the conversation

Small models around the conversation

Intent, entailment, embeddings, segmentation and voice activity detection run on self-hosted models, most under one billion parameters, instead of a frontier model.

03

Speech recognition under two billion parameters

Speech recognition under two billion parameters

Speech recognition under two billion parameters

Live transcription runs on speech recognition models we host on our own GPUs, sized between 0.6 and 1.7 billion parameters.

04

Speech that is cached, not re-synthesized

Speech that is cached, not re-synthesized

Speech that is cached, not re-synthesized

Identical agent utterances are served from a voice cache instead of being synthesized again by the text-to-speech provider.

05

Optimized inference, not just smaller weights

Optimized inference, not just smaller weights

Optimized inference, not just smaller weights

Self-hosted models run in half precision with ONNX Runtime paths. Model instances are pooled across calls, not reloaded for each one.

06

Single-GPU instances that follow call volume

Single-GPU instances that follow call volume

Single-GPU instances that follow call volume

Inference runs on single-GPU cloud instances that scale with live calls. Dispatch is capacity-aware, so we never dial more calls than we can serve.

07

Bounded conversations

Bounded conversations

Bounded conversations

Every call has a hard maximum duration and a silence timeout. A stalled or runaway session cannot keep burning compute.

A single backlit green leaf behind fluted glass

Hosting

On AWS, in regions AWS reports as renewable-matched.

On AWS, in regions AWS reports as renewable-matched.

On AWS, in regions AWS reports as renewable-matched.

Our regions, US East and Frankfurt, are on AWS’s 2025 list of regions whose electricity was matched 100 percent with renewable sources. AWS reports a global power usage effectiveness of 1.14.

Our regions, US East and Frankfurt, are on AWS’s 2025 list of regions whose electricity was matched 100 percent with renewable sources. AWS reports a global power usage effectiveness of 1.14.

Our regions, US East and Frankfurt, are on AWS’s 2025 list of regions whose electricity was matched 100 percent with renewable sources. AWS reports a global power usage effectiveness of 1.14.

AWS’s figures, on an annual-matching basis.

What the research says

Smaller models use far less energy. The evidence is public.

Smaller models use far less energy. The evidence is public.

Smaller models use far less energy. The evidence is public.

~100×

~100×

~100×

Size drives energy per token

Models in the 70-billion-parameter class consume on the order of 100 times more energy per token than small models.

The Price of Prompting, arXiv 2407.16893

70×

70×

70×

Spread on one short prompt

Across 30 commercial models, a short prompt ranged from about 0.1 Wh on a small model to about 7 Wh on a reasoning model.

How Hungry is AI?, arXiv 2505.09598. Third-party estimates.

90%

90%

90%

Task-specific beats general

Small models tailored to a task can cut energy use by up to 90 percent compared with one large general-purpose model.

UNESCO and UCL, July 2025

0.24 Wh

0.24 Wh

0.24 Wh

A median prompt, disclosed

Google reports a median Gemini text prompt at 0.24 Wh and 0.03 gCO2e. Mistral’s lifecycle assessment puts a 400-token response at 1.14 gCO2e.

Google, Aug 2025 (self-reported). Mistral with ADEME, July 2025.

None of these figures are ours. We cite them for direction, and to explain why model size is the lever we pull first.

One water droplet on a green shoot behind fluted glass

What we are working toward

Numbers before adjectives.

Numbers before adjectives.

Numbers before adjectives.

Things we can measure with tools we already run, and one direction we have chosen. This page changes when each has a number behind it.

Things we can measure with tools we already run, and one direction we have chosen. This page changes when each has a number behind it.

Things we can measure with tools we already run, and one direction we have chosen. This page changes when each has a number behind it.

01

Energy per call minute

Energy per call minute

Energy per call minute

Measure GPU energy on our own inference per minute of conversation, and publish the method alongside the figure.

02

Hosting footprint by region

Hosting footprint by region

Hosting footprint by region

Report the emissions our cloud provider attributes to our workloads, per region, using its customer carbon reporting.

03

Share of turns without the LLM

Share of turns without the LLM

Share of turns without the LLM

Our analytics already tag every scripted line as spoken without a language model call. We will publish that share of turns, and how it moves.

04

More of the dialogue on models we run

More of the dialogue on models we run

More of the dialogue on models we run

Move dialogue inference toward fine-tuned open models served on our own GPUs. Chosen, not done. We will say when it is.

Ask us how a call is served.

Ask us how a call is served.

Ask us how a call is served.

We will walk your team through the models, the hosting, and what we measure.

We will walk your team through the models, the hosting, and what we measure.

We will walk your team through the models, the hosting, and what we measure.

Contact Sales

© Harmony. A monday.com company.