---
title: "Local Models Tested in Voice Calls | Micdrop"
description: "Models tested in a Micdrop voice call, with their measurements and the reason each one is not in the recommended stack."
url: "https://micdrop.dev/docs/local-models/explorations"
---

# Local Models Tested in Voice Calls

Each page describes one model tested in a Micdrop call, as the agent, the transcription or the voice. A voice assistant needs things that general benchmarks do not measure: a short wait before the first word, speech generated faster than it plays, and tool calls in the right format during a conversation.

The tests run with the demo in `examples/advanced`. Its `server/src/benchmarks` folder has one script per model, so each result can be reproduced on another machine or with a newer release.

Model

Tested as

Why it was left out

[MiniCPM5-2B](/docs/local-models/explorations/minicpm)

The agent

Twice as fast as Qwen3 4B, but English only in practice

[MiniCPM-V 4.7](/docs/local-models/explorations/minicpm-v)

The agent

35B-A3B reliable but 15.5 GB, 1B never calls a tool

[Mistral 7B](/docs/local-models/explorations/mistral-7b)

The agent

Slower than Qwen3 4B, and its tool calls fail

[Voxtral Mini 3B](/docs/local-models/explorations/voxtral-stt)

Speech to text

More accurate than Whisper `base`, 35 times slower

[Voxtral TTS 4B](/docs/local-models/explorations/voxtral-tts)

The voice

Generates slower than real time

[AuK and AuK-Flash](/docs/local-models/explorations/auk)

The voice

Nine times slower than real time, even with a memory fix

All measurements were made on the machine described on the [latency and memory](/docs/local-models/performance) page. The results apply to that machine, not to the models in general. A server with a large GPU, or a faster runtime for the same weights, can make any of them usable.

MiniCPM5-2B is the exception. It is not in the recommended stack, but it is recommended for an English call on a machine with little memory, as the [local LLM page](/docs/local-models/agent) explains.
