watch Google Gemini · Research

Gemini 3.8 TTS Pricing Doubles on Jan 1, 2027: What Voice-App Builders Should Budget

Data graphic: dumbbell chart of Gemini 3.8 output prices per 1M audio tokens, Flash TTS $9.00 to $18.00 and Flash-Lite TTS $6.00 to $12.00, when intro rates end on December 31, 2026.
TM

Exploit intelligence researcher · Updated Oct 2, 2026, 3:27 PM EDT

Gemini 3.8 Flash TTS and Flash-Lite TTS launched at intro rates that double on January 1, 2027. What voice apps pay now, and what to budget.

Google's Gemini API pricing page lists Gemini 3.8 Flash TTS and Gemini 3.8 Flash-Lite TTS at introductory paid rates through December 31, 2026. From January 1, 2027 the listed rates double. Any voice product that ships on these models this quarter is therefore budgeting against a price that has an expiry date.

The rates

Both models went generally available on September 22, 2026 (gemini-3.8-flash-tts and gemini-3.8-flash-lite-tts). Flash-Lite TTS is positioned as the replacement for gemini-3.1-flash-tts-preview. Per 1M tokens, paid tier, standard:

ModelInput (text), nowOutput (audio), nowInput from 2027-01-01Output from 2027-01-01
Flash TTS$0.50$9.00$1.00$18.00
Flash-Lite TTS$0.50$6.00$1.00$12.00

Context caching is $0.125 per 1M tokens now and $0.25 from January 1. Batch pricing is listed at a 50% reduction on standard rates. Both TTS models list a free tier with free input and output.

Compared with Gemini 3.8 Live

Gemini 3.8 Live, released September 15, 2026, is priced for real-time dialogue: $0.75 per 1M text input tokens and $4.50 text output; audio is $3.00 per 1M input tokens (or $0.005 per minute) and $12.00 per 1M output tokens (or $0.018 per minute). It also lists a free tier. Unlike the TTS rows, the Live section carries no "through December 31, 2026" language on the page we read, so we state only its current rates. On the free tier, the page marks content as used to improve Google's products ("Yes", versus "No" on paid), which matters for any real user audio.

Worked cost example

Assumptions, all ours: 100,000 minutes of generated speech per month; Google's pricing page states "Audio tokens correspond to 25 tokens per second of audio" for the TTS models, so 1,500 tokens per minute. We also assume 250 text tokens of script per minute of speech, which is a guess.

  • Output: 100,000 x 1,500 = 150M audio tokens. Input: 25M text tokens.
  • Flash TTS: $1,350 + $12.50 = about $1,363 per month now; about $2,725 from January 1.
  • Flash-Lite TTS: $900 + $12.50 = about $913 now; about $1,825 from January 1.
  • Live, output audio only, at the per-minute rate: 100,000 x $0.018 = $1,800 (that rate matches $12 per 1M at 25 tokens per second). Audio input adds $500 if the user speaks for the same minutes.

Live is not a like-for-like substitute: it handles two-way conversation, while TTS only generates speech from text you supply (and you still pay for the LLM that writes it). Google also lists TTS output as $0.00225 per 10 seconds of audio for Flash and $0.0015 for Flash-Lite through 2026, doubling after.

What to budget

  • Model 2027 on the doubled rates, not the intro ones. Flash TTS output goes from $9 to $18, Flash-Lite from $6 to $12.
  • Flash-Lite is one third cheaper than Flash on output; test whether its voice quality meets your bar before committing.
  • Use batch (50% off) for anything non-interactive, such as audiobooks or pre-rendered prompts, and cache repeated prefixes.
  • Cache or store generated audio for repeated phrases.
  • Do not send customer audio through the free tier.
  • Re-read the pricing page before January 1 in case Live pricing changes.

Sources

Keep reading

All latest →
  1. watchResearchXing4.0-29B-A4B: China Telecom's Open Coding Model Claims SWE-bench 75 on One GPU6 min
  2. watchResearchGitHub Copilot Drops Six Models on Oct 19: What Replaces GPT-5.5, GPT-5.4 and Grok 4.57 min
  3. elevatedResearchClaude Sonnet 4.5 Retires Nov 30: Six Requests That Return a 400 on Sonnet 5.58 min
  4. watchResearchGemini 4 Argon vs GPT-6 Astra vs Claude Opus 5.5: Is Google's New Model Actually Better?6 min
  5. stableResearchLing 3.1 Flash vs GLM 5.3 Flash vs Qwen 3.8 Flash Next: Three Bets on Cheap Agent Models6 min
  6. elevatedResearchLing-3.1-flash Is Ant Group's Best Flash Model Yet, and It Scores 87.9 on CyberGym6 min