Phi 4 Multimodal Instruct

Feb 2025
Phi 4 Multimodal Instruct
Compare in
on
$0.05per 1M billed tokens
5Open ASR
Feb 2025released
en, es, fr +20Languages

Ranking?

Full ranking →
1Zoom Scribe V2 Pro6.12ASR K16.33Multilingual6.34Speech v3.16.45Zoom Scribe v16.9
···
31Smallest AI Pulse9.132Phi 4 Multimodal Instruct9.133Muse Voice Transcribe9.2

Run it yourself

estimate
Fits on✓ 8 GB (phone, laptop) · 8-bit✓ 16 GB · full quality✓ 24 GB GPU · full quality✓ 48 GB GPU · full quality✓ 80 GB GPU · full quality
LicenceMITfree for commercial use

Benchmarks

in Speech & Voice
Open ASR · average WER5#7
Open ASR · RTFx162.5#38
Open ASR · AMI Cleaned WER9.1#32
Open ASR · Earnings-22 Cleaned AA Chunked WER7.1#30
Open ASR · GigaSpeech Cleaned WER7.8#21
Open ASR · LibriSpeech Clean WER1.4#30
Open ASR · LibriSpeech Other WER3.5#38
Open ASR · SPGISpeech WER2.7#34
Open ASR · Voice Arena Monsoon WER4.5#36
Open ASR · VoxPopuli AA Cleaned WER4.2#40
Open ASR · AMI WER10.7#18
Open ASR · Earnings-22 WER9.4#14
Open ASR · GigaSpeech WER7.9#8
Open ASR · VoxPopuli WER5.8#13
Open ASR · French RTFx127.7#18
Open ASR · French FLEURS WER4.1#13
Open ASR · French MCV WER6.9#18
Open ASR · French MLS WER4.1#19
Open ASR · German RTFx122.7#18
Open ASR · German FLEURS WER4#23
Open ASR · German MCV WER4.3#19
Open ASR · Spanish RTFx135.3#19
Open ASR · Spanish FLEURS WER3.2#22
Open ASR · Spanish MCV WER4#21
Open ASR · Spanish MLS WER4.1#27
Open ASR · Italian RTFx122.6#18
Open ASR · Italian FLEURS WER1.9#12
Open ASR · Italian MCV WER3.8#15
Open ASR · Italian MLS WER7.2#25
Open ASR · Portuguese RTFx118.5#20
Open ASR · Portuguese FLEURS WER3.4#10
Open ASR · Portuguese MLS WER6.9#21

FAQ

How much does Phi 4 Multimodal Instruct cost?›

$0.05 per 1M billed tokens.

Is Phi 4 Multimodal Instruct free to use?›

There is no free API; it costs $0.05 per 1M billed tokens. The weights are open, so you can run it yourself.

How good is Phi 4 Multimodal Instruct, and what is it best at?›

#7 of 61 in Speech to text.

Is Phi 4 Multimodal Instruct better than Smallest AI Pulse?›

On Open ASR · AMI Cleaned WER, Phi 4 Multimodal Instruct scores 9.1 vs 9.1 — about the same. It is 900% more expensive.

When was Phi 4 Multimodal Instruct released? Is there a newer version?›

Released 2025-02-24. It is the current version in its line.

Is Phi 4 Multimodal Instruct open source?›

Yes, the weights are open (MIT licence).

Can I run Phi 4 Multimodal Instruct locally?›

The weights are open and can be self-hosted; hardware needs are not confirmed yet.

What is Phi 4 Multimodal Instruct?›

Phi 4 Multimodal Instruct is an open-weights speech & voice model from Microsoft, released Feb 2025.

What are the best alternatives to Phi 4 Multimodal Instruct?›

Multilingual (better), Granite Speech 4.1 2B Nar (open weights).