Light week: no verified rate cut

Week 35: transcription costs and migration deadlines

No verified subscription or API rate cut landed in Week 35. The reporting window is August 24-30, 2026. The real wallet work is planning two transcription migrations and one same-price Google endpoint move.

💰 Your Wallet This Week

#1 OpenAI standard file transcription: $6.00 becomes $4.50 per 1,000 minutes, saving $1.50 (25%).
#2 OpenAI mini transcription: $3.00 becomes $4.50 per 1,000 minutes, adding $1.50 (50%).
#3 Google's new file model: 1,000 audio minutes carries $1.92 Standard input or $0.96 Batch input, before text output.

Price Change Summary

Tool	What Changed	Was	Now	Impact	Who It Affects
🧠 OpenAI transcription	Four models received a February 2027 shutdown date	$0.003-$0.006/min for three file models	$0.0045/min file or $0.017/min live	Mixed	API users on the four retiring models
🎤 Gemini Transcribe	Two transcription models reached GA with new rate cards	Not generally available	From $0.50/M audio input in Batch	New cost	File and live transcription developers
🎥 Gemini Omni Flash	Stable endpoint launched; preview shutdown scheduled	$1.50 / $9 / $17.50 per 1M	Same published rates	No change	Developers using the preview endpoint

Previous report: Week 34, August 17-23, 2026. Browse the full briefing archive.

🧠

OpenAI set a February 2027 transcription deadline

OpenAI announced the deprecation on August 26, 2026. The affected models are whisper-1, gpt-4o-transcribe, gpt-4o-mini-transcribe, and gpt-4o-transcribe-diarize. They shut down on February 26, 2027.

OpenAI points customers to gpt-transcribe or gpt-live-transcribe. gpt-transcribe costs $0.0045 per minute. gpt-live-transcribe costs $0.017 per minute.

For file workloads moving from whisper-1 or gpt-4o-transcribe at $0.006/min, $6.00 becomes $4.50 per 1,000 minutes, saving $1.50 (25%). For gpt-4o-mini-transcribe at $0.003/min, $3.00 becomes $4.50 per 1,000 minutes, adding $1.50 (50%).

gpt-4o-transcribe-diarize uses token pricing of $2.50 input and $10.00 output per million tokens; no per-minute comparison is published here.

⚠️ Who This Affects

API teams using any of the four retiring transcription models. Inventory model IDs now. File and live replacements have different cost and latency profiles.

Official sources checked August 31, 2026: OpenAI's dated API deprecations establishes the retirement schedule; current API pricing and the official gpt-transcribe and gpt-live-transcribe model pages support current replacement rates. Current pages are not used alone to prove the historical deadline.

🎤

Google launched dedicated file and live transcription rates

Google released Gemini 3.5 Transcribe and Gemini 3.5 Transcribe Live as generally available on August 26, 2026.

Gemini 3.5 Transcribe Standard: $0.30 text input, $1.00 audio input, and $2.50 text output per million tokens. Batch: $0.15 text input, $0.50 audio input, and $1.25 text output per million tokens.

Gemini 3.5 Transcribe Live: $0.50 text input, $3.00 audio or video input, and $2.00 text output per million tokens.

At 32 audio tokens per second, 1,000 minutes equals 1.92 million audio tokens: $1.92 Standard input or $0.96 Batch input, before text output. Live audio or video input is $5.76 for the same token count, before text output.

This is a new rate card, not a verified reduction from an older Google price.

🎤 Who This Affects

Developers pricing new file or live transcription workloads. Output tokens are billed separately, so do not treat the input-only example as the final invoice.

Official sources checked August 31, 2026: Google's dated Gemini API changelog establishes the launch; the file-transcription and live-transcription pricing sections support the published launch rates. No earlier Google rate is inferred.

🎥

Gemini Omni moved to GA without a listed price change

Google released gemini-omni-1.1-flash as GA on August 27, 2026. gemini-omni-flash-preview shuts down September 30, 2026.

Both publish $1.50 input, $9.00 text output, and $17.50 video output per million tokens.

Google bills 720p output at 5,792 tokens per second. A 10-second 720p output is about $1.01 before input charges on either endpoint, so the migration itself changes the listed price by $0.00.

🎥 Who This Affects

Developers still calling the preview endpoint. Move before September 30 for availability, not for a lower published rate.

Official sources checked August 31, 2026: Google's dated Gemini API changelog, deprecations table, model page, and current pricing section. The article reports no listed price change; it does not use today's rate card to manufacture one.

🔍 Checked and excluded

These items were in the Week 35 source sweep but did not clear the price-change bar.

xAI Grok Bot: xAI's August 11 and August 26 Grok Bot posts list the same eligible plans, so no expansion claim is published.
Anthropic Memory: the August 25 release note changes feature availability across plans, not a verified subscription price.
Canva nonprofits: the August 24 post is a participation milestone, not evidence of a new weekly price policy.
DeepSeek: an August 28 page timestamp proves retrieval recency, not a before value or a price-change date.

🔍 Switch Watch

Three actions based only on this week's verified pricing evidence.

Action Required

Find retiring OpenAI model IDs

Search production traffic for the four deprecated model names before the February 26, 2027 shutdown.

Opportunity

Benchmark file transcription now

Standard users may save $1.50 per 1,000 minutes. Mini users may pay the same amount more.

Info

Move off Omni preview

The published rate difference is $0.00. Migrate for endpoint continuity before September 30.

Method: official changelogs establish dates and lifecycle changes. Official pricing and model pages establish current amounts and units. Current pages alone are not treated as proof of a historical price change.