Skip to content

Whisper API Pricing: $0.006 a Minute Until the Feb 2027 Shutdown

Short answer. The OpenAI Whisper API (model whisper-1) is listed at $0.006 per minute of audio, about $0.36 an hour, in the estimated cost column of OpenAI's pricing page. It is already marked deprecated: OpenAI's deprecations page says whisper-1 shuts down on February 26, 2027, with gpt-transcribe or gpt-live-transcribe as the replacement. gpt-transcribe is listed at $0.0045 a minute, 25% less than Whisper.

Whisper API pricing: our arithmetic on OpenAI estimated costs and Groq listed prices for 3,000 minutes of audio

This guide is for founders, developers and small teams that run Whisper on recorded sales calls, voicemails, meetings or support calls. It covers every transcription price on OpenAI's pricing page as read on November 17, 2026, what the February 2027 shutdown changes, what a typical monthly volume costs on each model, and what Groq charges to run the open Whisper model. Worked examples are our own arithmetic on published prices.

Key numbers

ItemNumber
Whisper (whisper-1) (OpenAI's estimated cost; $0.36 an hour)$0.006 a minute
Whisper shutdown (announced Aug 26, 2026)Feb 26, 2027
gpt-transcribe (25% less than Whisper)$0.0045 a minute
gpt-4o-mini-transcribe (also deprecated; gpt-transcribe is 50% more)$0.003 a minute
gpt-live-transcribe (streaming audio)$0.017 a minute
File size limit (OpenAI file transcription)25 MB
Groq whisper-large-v3 (10 second minimum per request)$0.111 an hour
Groq whisper-large-v3-turbo (no translation support)$0.04 an hour
3,000 minutes on Whisper (our arithmetic, hypothetical volume)$18.00

Source: OpenAI API pricing, deprecations, changelog and file transcription guide; Groq speech to text docs; read November 17, 2026.

Whisper API price per minute and per hour

OpenAI's pricing page lists eight transcription and translation models. Whisper and the gpt-4o models are flagged "Deprecated". The per hour column is our arithmetic (per minute times 60).

ModelUseListed cost per minutePer hourStatus
whisper-1 (Whisper)Transcription$0.006$0.36Deprecated, shuts down Feb 26, 2027
gpt-transcribeTranscription$0.0045$0.27Not deprecated; recommended
gpt-4o-transcribeTranscription$0.006$0.36Deprecated, shuts down Feb 26, 2027
gpt-4o-mini-transcribeTranscription$0.003$0.18Deprecated, shuts down Feb 26, 2027
gpt-4o-transcribe-diarizeTranscription with speaker labels$0.006$0.36Deprecated, shuts down Feb 26, 2027
gpt-live-transcribeLive (streaming) transcription$0.017$1.02Not deprecated
gpt-realtime-whisperLive (streaming) transcription$0.017$1.02Not deprecated
gpt-realtime-translateLive translation$0.034$2.04Not deprecated

Two details matter when you read that table. First, OpenAI labels the per minute figures "Estimated cost". For Whisper and gpt-transcribe the page shows no token rates, only the per minute estimate. Second, the three deprecated gpt-4o models also carry token rates: $2.50 input and $10.00 output per million tokens for gpt-4o-transcribe and the diarize model, and $1.25 and $5.00 for gpt-4o-mini-transcribe. Those token rates sit alongside the per minute estimate.

The February 26, 2027 shutdown

OpenAI announced on August 26, 2026 "the deprecation of whisper-1, gpt-4o-transcribe, gpt-4o-mini-transcribe, and gpt-4o-transcribe-diarize. These models will shut down on February 26, 2027." The same date applies to the /v1/audio/translations endpoint, according to OpenAI's file transcription guide.

Counting from November 17, 2026, that leaves 101 days. On OpenAI's deprecations page, "Deprecated" means "You can still use it until its shutdown date", and "Shutdown" means "The model no longer accepts requests through the standard API." So a Whisper integration keeps working until February 26, 2027 (listed today at $0.006 a minute), and then its requests stop being accepted. The page also says "Impacted customers will always be notified by email and in our documentation."

The deprecations page names gpt-live-transcribe or gpt-transcribe as the replacement. The changelog of July 28, 2026, when both were released, describes gpt-transcribe for "accurate file transcription" and gpt-live-transcribe for "low-latency streaming transcription". OpenAI says both "support free-form transcription context, keyword hints, and multiple expected input languages."

What moving off Whisper costs you

The price effect depends on which deprecated model you run today. Our arithmetic on the listed estimates:

  • From whisper-1 to gpt-transcribe: $0.006 to $0.0045 a minute, 25% lower.
  • From gpt-4o-transcribe or gpt-4o-transcribe-diarize to gpt-transcribe: the same 25% drop, since both deprecated models list $0.006.
  • From gpt-4o-mini-transcribe to gpt-transcribe: $0.003 to $0.0045 a minute, 50% higher. Among the deprecated file models, this is the one move that raises the bill.
  • From any file model to gpt-live-transcribe: $0.017 a minute, about 3.8 times gpt-transcribe. Use it only when you need text while the audio is still arriving.

For recorded calls, OpenAI's guide recommends gpt-transcribe; check the features below before you switch. Live transcription is a different job, and the guide sends audio "still arriving from a microphone, call, or media stream" to Realtime transcription.

Cost of 3,000 minutes of audio by model
Our arithmetic on OpenAI estimated costs and Groq listed prices. Source: developers.openai.com and console.groq.com, November 2026

What a month of call transcripts costs

A hypothetical sales team records 500 calls a month that average 6 minutes each: 3,000 minutes of audio. At the listed estimates, that month costs:

Model3,000 minutes1,000 hours
whisper-1$18.00$360
gpt-4o-transcribe$18.00$360
gpt-transcribe$13.50$270
gpt-4o-mini-transcribe$9.00$180
gpt-live-transcribe$51.00$1,020

At this volume the switch from Whisper to gpt-transcribe saves $4.50 a month, so price is not the reason to plan the move: the shutdown date is. At 1,000 hours a month the gap is $90. If you also run a model over each transcript to score a call, write CRM notes or draft a follow-up, that is billed separately in tokens. If transcripts feed your CRM and follow-ups, our guide to sales automation software covers the tools that pick them up.

What gpt-transcribe does and what the guide still shows on Whisper

OpenAI's file transcription guide says: "Start with gpt-transcribe. This is the recommended model for transcribing recorded speech in its original language." It returns the transcript and the detected languages as JSON, and accepts a prompt, keywords and expected languages "to improve transcription of domain terms and multilingual audio".

The same guide adds: "The specialized examples for speaker labels, word timestamps, and subtitle formats below use deprecated models." Specifically:

  • Word or segment timestamps: the guide's example uses whisper-1, available "until February 26, 2027".
  • Speaker labels: the guide points existing integrations to gpt-4o-transcribe-diarize, also until February 26, 2027.
  • Translation to English: the /v1/audio/translations endpoint uses whisper-1 and shuts down the same day. The guide says to "Migrate to Realtime translation with gpt-realtime-translate", a streaming session listed at $0.034 a minute.

If your workflow depends on timestamps for captions, speaker labels for call coaching, or translation of recorded files, test what gpt-transcribe returns before February. We did not find a statement on that page that gpt-transcribe offers those outputs.

File size and format limits

The guide sets the same limits for file transcription: "Files can be up to 25 MB. Supported input formats include flac, mp3, mp4, mpeg, mpga, m4a, ogg, wav, and webm." For longer recordings it says to "use a compressed audio format or split the file into chunks of 25 MB or less", and to avoid splitting in the middle of a sentence.

Is Whisper free? The open model and hosted options

The Whisper API is not free: OpenAI's pricing page lists an estimated $0.006 per minute. Our guide to the OpenAI API free tier covers what a new account gets. The Whisper model itself is open: on GitHub, OpenAI says "Whisper's code and model weights are released under the MIT License." That repository lists six model sizes, from tiny (39 million parameters, about 1 GB of VRAM) to large (1,550 million parameters, about 10 GB), and turbo (809 million, about 6 GB). Running it yourself means paying for the hardware instead of a per minute fee, and the February 2027 shutdown applies to OpenAI's API, not to those published weights.

Groq, for example, hosts the open model. Its speech to text docs list whisper-large-v3 at $0.111 an hour and whisper-large-v3-turbo at $0.04 an hour. Per minute, that is about $0.00185 and $0.00067, roughly a third and a ninth of OpenAI's Whisper estimate. Groq lists translation support for large-v3 but not for turbo, and bills a minimum of 10 seconds per request: "If you submit a request less than this, you will still be billed for 10 seconds." For a wider price comparison across transcription vendors, read our page on AI transcription cost per hour.

A short migration checklist

  1. Search your code for whisper-1, the three gpt-4o transcribe models and /v1/audio/translations.
  2. Run a sample of real recordings through gpt-transcribe with your product names in keywords, and compare the output with your Whisper transcripts.
  3. Check any step that reads timestamps, speaker labels or subtitle formats, since the guide's examples for those use deprecated models.
  4. Recompute your monthly cost at $0.0045 a minute, or at $0.017 if you move to live transcription.
  5. Ship the switch well before February 26, 2027. Our page on the GPT-4o API shutdown lists the other OpenAI models retiring in the same window.

If you record calls to transcribe them, the consent rules for recording come first; our state by state guide to call recording laws covers them. This is general information, not legal advice.

How imisofts can help

imisofts builds AI voice agents, call workflows and automation; see our pages on AI calling for home services and our AI automation agency. If you are planning a transcription workflow, we can talk it through with you. For the agents that answer calls themselves, see our comparison of the best AI voice agents.

What we could not verify

  • Whether OpenAI bills Whisper to the second or rounds to the minute. The pricing page shows only an estimated cost per minute.
  • Whether gpt-transcribe returns word timestamps, speaker labels or subtitle formats. The guide shows those only on deprecated models.
  • Groq's main pricing page, which redirected to its home page when we checked; Groq figures above come from its speech to text docs.

Planning a call transcription workflow? Talk to us

We design, build, and run it for you, integrated with the tools you already use. Free audit in 24 hours.

Get Your Free Audit

Frequently Asked Questions

OpenAI's pricing page lists Whisper (whisper-1) at an estimated $0.006 per minute of audio, which is $0.36 an hour, as read on November 17, 2026. A hypothetical 3,000 minutes of call recordings a month comes to $18.00.

Yes. OpenAI announced on August 26, 2026 that whisper-1, gpt-4o-transcribe, gpt-4o-mini-transcribe and gpt-4o-transcribe-diarize shut down on February 26, 2027. Its file transcription guide says the /v1/audio/translations endpoint shuts down the same day. The deprecations page names gpt-transcribe or gpt-live-transcribe as the replacement.

For recorded files, gpt-transcribe, listed at an estimated $0.0045 a minute, 25% less than Whisper. For streaming audio, gpt-live-transcribe at $0.017 a minute. Both were released on July 28, 2026.

No. The API is billed per minute. The Whisper model itself is open: OpenAI released its code and weights under the MIT License on GitHub, so you can run it on your own hardware and pay for compute instead of a per minute fee.

Groq's docs list whisper-large-v3 at $0.111 an hour and whisper-large-v3-turbo at $0.04 an hour, about $0.00185 and $0.00067 a minute. OpenAI lists Whisper at $0.006 a minute. Groq bills at least 10 seconds per request.

OpenAI's file transcription guide says files can be up to 25 MB, in flac, mp3, mp4, mpeg, mpga, m4a, ogg, wav or webm. Longer recordings need compression or chunks of 25 MB or less.

Free Strategy Audit

Ready to put this to work?

Join 200+ businesses already scaling with AI and automation. Get your free audit and a custom roadmap within 48 hours.

Website & marketing performance analysis
AI & automation opportunity mapping
Custom growth roadmap with ROI estimates
Delivered within 48 hours, 100% free
200+
Clients served
48hr
Turnaround
100%
Free, no strings

Get Your Free Audit

Takes 30 seconds. No credit card required.

Prefer to chat?

WhatsApp us