Email me

LLM prices / Price history

AI price history · 17 families · checked Oct 23, 2026

AI model price history: are API prices falling?

Short answerIt depends on the family. In 8 of 17 model families, the newest model still on the pricing page costs less per chatbot reply than the oldest one; in 6 it costs more and in 3 it is the same. The biggest fall: Claude Haiku, from $2.80 per 1,000 replies on Claude Haiku 3.5 to $0.35 on Claude Haiku 5.5 (-88%). Pricier now: GPT mini (+529%), Gemini Flash-Lite (+368%), GPT nano (+158%), Gemini Flash (+81%), GPT Pro (+66%), Gemini Pro (+33%). Between our 2 dated checks of the pricing pages (Oct 17, 2026, Oct 23, 2026) no existing model changed price, and providers have announced 3 future price changes.

17model families, 61 models
8 of 17families cheaper now than their oldest listed model
-88%biggest fall (Claude Haiku)
2dated checks, first Oct 17, 2026
Bar chart of the biggest generation-to-generation price drops in AI model families
Proof Engine render from the same table. Cost of 1,000 chatbot replies at list price.
Every familyEvery step, biggest drop firstWhat we record over timeDownload CSVRaw checks CSVEmbed a live price

Each family, oldest model to newest

A family is one product line from one provider (Claude Opus, GPT mini, Gemini Flash). The bars price one chatbot reply (1,500 tokens in, 400 out) times 1,000, at each model's list price on the provider's page today. Click any model for every price tier and its own chart.

Claude Haiku3 models, oldest to newest: -88%

  1. Claude Haiku 3.5retired$2.80$0.8 in / $4 out per 1M tokens
  2. Claude Haiku 4.5legacy$3.50$1 in / $5 out per 1M tokens (+25% vs the one before)
  3. Claude Haiku 5.5$0.35$0.1 in / $0.5 out per 1M tokens (-90% vs the one before)

OpenAI o-series2 models, oldest to newest: -87%

  1. o1shut down$46.50$15 in / $60 out per 1M tokens
  2. o3deprecated$6.20$2 in / $8 out per 1M tokens (-87% vs the one before)

OpenAI o pro2 models, oldest to newest: -87%

  1. o1-proshut down$465$150 in / $600 out per 1M tokens
  2. o3-prodeprecated$62.00$20 in / $80 out per 1M tokens (-87% vs the one before)

Claude Opus8 models, oldest to newest: -73%

  1. Claude Opus 4retired$52.50$15 in / $75 out per 1M tokens
  2. Claude Opus 4.1retired$52.50$15 in / $75 out per 1M tokens (+0% vs the one before)
  3. Claude Opus 4.5legacy$17.50$5 in / $25 out per 1M tokens (-67% vs the one before)
  4. Claude Opus 4.6legacy$17.50$5 in / $25 out per 1M tokens (+0% vs the one before)
  5. Claude Opus 4.7legacy$17.50$5 in / $25 out per 1M tokens (+0% vs the one before)
  6. Claude Opus 4.8legacy$17.50$5 in / $25 out per 1M tokens (+0% vs the one before)
  7. Claude Opus 5legacy$17.50$5 in / $25 out per 1M tokens (+0% vs the one before)
  8. Claude Opus 5.5$14.00$4 in / $20 out per 1M tokens (-20% vs the one before)

GPT, numbered line10 models, oldest to newest: -72%

  1. GPT-4 (0613)no spec page$69.00$30 in / $60 out per 1M tokens
  2. GPT-4 Turbo (2024-04-09)no spec page$27.00$10 in / $30 out per 1M tokens (-61% vs the one before)
  3. GPT-4o (2024-05-13)no spec page$13.50$5 in / $15 out per 1M tokens (-50% vs the one before)
  4. GPT-4odeprecated$7.75$2.5 in / $10 out per 1M tokens (-43% vs the one before)
  5. GPT-4.1deprecated$6.20$2 in / $8 out per 1M tokens (-20% vs the one before)
  6. GPT-5deprecated$5.88$1.25 in / $10 out per 1M tokens (-5% vs the one before)
  7. GPT-5.1deprecated$5.88$1.25 in / $10 out per 1M tokens (+0% vs the one before)
  8. GPT-5.2$8.22$1.75 in / $14 out per 1M tokens (+40% vs the one before)
  9. GPT-5.4$9.75$2.5 in / $15 out per 1M tokens (+19% vs the one before)
  10. GPT-5.5$19.50$5 in / $30 out per 1M tokens (+100% vs the one before)

GPT Luna2 models, oldest to newest: -55%

  1. GPT-5.6 Luna$0.78$0.2 in / $1.2 out per 1M tokens
  2. GPT-6 Luna$0.35$0.1 in / $0.5 out per 1M tokens (-55% vs the one before)

GPT Sol3 models, oldest to newest: -50%

  1. GPT-5.6 Sol$14.00$4 in / $20 out per 1M tokens
  2. GPT-6 Sol$7.00$2 in / $10 out per 1M tokens (-50% vs the one before)
  3. GPT-6.1 Sol$7.00$2 in / $10 out per 1M tokens (+0% vs the one before)

Claude Sonnet5 models, oldest to newest: -33%

  1. Claude Sonnet 4retired$10.50$3 in / $15 out per 1M tokens
  2. Claude Sonnet 4.5deprecated$10.50$3 in / $15 out per 1M tokens (+0% vs the one before)
  3. Claude Sonnet 4.6legacy$10.50$3 in / $15 out per 1M tokens (+0% vs the one before)
  4. Claude Sonnet 5legacy$7.00$2 in / $10 out per 1M tokens (-33% vs the one before)
  5. Claude Sonnet 5.5$7.00$2 in / $10 out per 1M tokens (+0% vs the one before)

Claude Fable2 models, oldest to newest: +0%

  1. Claude Fable 5legacy$35.00$10 in / $50 out per 1M tokens
  2. Claude Fable 5.1$35.00$10 in / $50 out per 1M tokens (+0% vs the one before)

GPT Cyber2 models, oldest to newest: +0%

  1. GPT-5.5 Cyberno spec page$48.75$12.5 in / $75 out per 1M tokens
  2. GPT-5.6 Cyber$48.75$12.5 in / $75 out per 1M tokens (+0% vs the one before)

OpenAI o mini2 models, oldest to newest: +0%

  1. o3-minishut down$3.41$1.1 in / $4.4 out per 1M tokens
  2. o4-minishut down$3.41$1.1 in / $4.4 out per 1M tokens (+0% vs the one before)

Gemini Pro2 models, oldest to newest: +33%

  1. Gemini 2.5 Prolimited access$5.88$1.25 in / $10 out per 1M tokens
  2. Gemini 3.1 Pro Previewpreview$7.80$2 in / $12 out per 1M tokens (+33% vs the one before)

GPT Pro4 models, oldest to newest: +66%

  1. GPT-5 Prodeprecated$70.50$15 in / $120 out per 1M tokens
  2. GPT-5.2 Pro$98.70$21 in / $168 out per 1M tokens (+40% vs the one before)
  3. GPT-5.4 Pro$117$30 in / $180 out per 1M tokens (+19% vs the one before)
  4. GPT-5.5 Pro$117$30 in / $180 out per 1M tokens (+0% vs the one before)

Gemini Flash4 models, oldest to newest: +81%

  1. Gemini 2.5 Flashlimited access$1.45$0.3 in / $2.5 out per 1M tokens
  2. Gemini 3 Flash Previewpreview$1.95$0.5 in / $3 out per 1M tokens (+34% vs the one before)
  3. Gemini 3.6 Flash$2.62$0.75 in / $3.75 out per 1M tokens (+35% vs the one before)
  4. Gemini 3.8 Flash$2.62$0.75 in / $3.75 out per 1M tokens (+0% vs the one before)

GPT nano3 models, oldest to newest: +158%

  1. GPT-4.1 nanoshut down$0.31$0.1 in / $0.4 out per 1M tokens
  2. GPT-5 nanodeprecated$0.23$0.05 in / $0.4 out per 1M tokens (-24% vs the one before)
  3. GPT-5.4 nanodeprecated$0.80$0.2 in / $1.25 out per 1M tokens (+240% vs the one before)

Gemini Flash-Lite3 models, oldest to newest: +368%

  1. Gemini 2.5 Flash-Litelimited access$0.31$0.1 in / $0.4 out per 1M tokens
  2. Gemini 3.1 Flash-Lite$0.97$0.25 in / $1.5 out per 1M tokens (+215% vs the one before)
  3. Gemini 3.5 Flash-Lite$1.45$0.3 in / $2.5 out per 1M tokens (+49% vs the one before)

GPT mini4 models, oldest to newest: +529%

  1. GPT-4o minideprecated$0.47$0.15 in / $0.6 out per 1M tokens
  2. GPT-4.1 minideprecated$1.24$0.4 in / $1.6 out per 1M tokens (+167% vs the one before)
  3. GPT-5 minideprecated$1.18$0.25 in / $2 out per 1M tokens (-5% vs the one before)
  4. GPT-5.4 mini$2.93$0.75 in / $4.5 out per 1M tokens (+149% vs the one before)

Every step, biggest drop first

FamilyFrom → toInput / 1MOutput / 1M1,000 chatbot repliesChange
Claude HaikuClaude Haiku 4.5 → Claude Haiku 5.5$1 → $0.1$5 → $0.5$3.50 → $0.35-90%
OpenAI o-serieso1 → o3$15 → $2$60 → $8$46.50 → $6.20-87%
OpenAI o proo1-pro → o3-pro$150 → $20$600 → $80$465 → $62.00-87%
Claude OpusClaude Opus 4.1 → Claude Opus 4.5$15 → $5$75 → $25$52.50 → $17.50-67%
GPT, numbered lineGPT-4 (0613) → GPT-4 Turbo (2024-04-09)$30 → $10$60 → $30$69.00 → $27.00-61%
GPT LunaGPT-5.6 Luna → GPT-6 Luna$0.2 → $0.1$1.2 → $0.5$0.78 → $0.35-55%
GPT, numbered lineGPT-4 Turbo (2024-04-09) → GPT-4o (2024-05-13)$10 → $5$30 → $15$27.00 → $13.50-50%
GPT SolGPT-5.6 Sol → GPT-6 Sol$4 → $2$20 → $10$14.00 → $7.00-50%
GPT, numbered lineGPT-4o (2024-05-13) → GPT-4o$5 → $2.5$15 → $10$13.50 → $7.75-43%
Claude SonnetClaude Sonnet 4.6 → Claude Sonnet 5$3 → $2$15 → $10$10.50 → $7.00-33%
GPT nanoGPT-4.1 nano → GPT-5 nano$0.1 → $0.05$0.4 → $0.4$0.31 → $0.23-24%
Claude OpusClaude Opus 5 → Claude Opus 5.5$5 → $4$25 → $20$17.50 → $14.00-20%
GPT, numbered lineGPT-4o → GPT-4.1$2.5 → $2$10 → $8$7.75 → $6.20-20%
GPT, numbered lineGPT-4.1 → GPT-5$2 → $1.25$8 → $10$6.20 → $5.88-5%
GPT miniGPT-4.1 mini → GPT-5 mini$0.4 → $0.25$1.6 → $2$1.24 → $1.18-5%
Claude OpusClaude Opus 4 → Claude Opus 4.1$15 → $15$75 → $75$52.50 → $52.50+0%
Claude OpusClaude Opus 4.5 → Claude Opus 4.6$5 → $5$25 → $25$17.50 → $17.50+0%
Claude OpusClaude Opus 4.6 → Claude Opus 4.7$5 → $5$25 → $25$17.50 → $17.50+0%
Claude OpusClaude Opus 4.7 → Claude Opus 4.8$5 → $5$25 → $25$17.50 → $17.50+0%
Claude OpusClaude Opus 4.8 → Claude Opus 5$5 → $5$25 → $25$17.50 → $17.50+0%
Claude SonnetClaude Sonnet 4 → Claude Sonnet 4.5$3 → $3$15 → $15$10.50 → $10.50+0%
Claude SonnetClaude Sonnet 4.5 → Claude Sonnet 4.6$3 → $3$15 → $15$10.50 → $10.50+0%
Claude SonnetClaude Sonnet 5 → Claude Sonnet 5.5$2 → $2$10 → $10$7.00 → $7.00+0%
Claude FableClaude Fable 5 → Claude Fable 5.1$10 → $10$50 → $50$35.00 → $35.00+0%
GPT, numbered lineGPT-5 → GPT-5.1$1.25 → $1.25$10 → $10$5.88 → $5.88+0%
GPT ProGPT-5.4 Pro → GPT-5.5 Pro$30 → $30$180 → $180$117 → $117+0%
GPT SolGPT-6 Sol → GPT-6.1 Sol$2 → $2$10 → $10$7.00 → $7.00+0%
GPT CyberGPT-5.5 Cyber → GPT-5.6 Cyber$12.5 → $12.5$75 → $75$48.75 → $48.75+0%
OpenAI o minio3-mini → o4-mini$1.1 → $1.1$4.4 → $4.4$3.41 → $3.41+0%
Gemini FlashGemini 3.6 Flash → Gemini 3.8 Flash$0.75 → $0.75$3.75 → $3.75$2.62 → $2.62+0%
GPT, numbered lineGPT-5.2 → GPT-5.4$1.75 → $2.5$14 → $15$8.22 → $9.75+19%
GPT ProGPT-5.2 Pro → GPT-5.4 Pro$21 → $30$168 → $180$98.70 → $117+19%
Claude HaikuClaude Haiku 3.5 → Claude Haiku 4.5$0.8 → $1$4 → $5$2.80 → $3.50+25%
Gemini ProGemini 2.5 Pro → Gemini 3.1 Pro Preview$1.25 → $2$10 → $12$5.88 → $7.80+33%
Gemini FlashGemini 2.5 Flash → Gemini 3 Flash Preview$0.3 → $0.5$2.5 → $3$1.45 → $1.95+34%
Gemini FlashGemini 3 Flash Preview → Gemini 3.6 Flash$0.5 → $0.75$3 → $3.75$1.95 → $2.62+35%
GPT, numbered lineGPT-5.1 → GPT-5.2$1.25 → $1.75$10 → $14$5.88 → $8.22+40%
GPT ProGPT-5 Pro → GPT-5.2 Pro$15 → $21$120 → $168$70.50 → $98.70+40%
Gemini Flash-LiteGemini 3.1 Flash-Lite → Gemini 3.5 Flash-Lite$0.25 → $0.3$1.5 → $2.5$0.97 → $1.45+49%
GPT, numbered lineGPT-5.4 → GPT-5.5$2.5 → $5$15 → $30$9.75 → $19.50+100%
GPT miniGPT-5 mini → GPT-5.4 mini$0.25 → $0.75$2 → $4.5$1.18 → $2.93+149%
GPT miniGPT-4o mini → GPT-4.1 mini$0.15 → $0.4$0.6 → $1.6$0.47 → $1.24+167%
Gemini Flash-LiteGemini 2.5 Flash-Lite → Gemini 3.1 Flash-Lite$0.1 → $0.25$0.4 → $1.5$0.31 → $0.97+215%
GPT nanoGPT-5 nano → GPT-5.4 nano$0.05 → $0.2$0.4 → $1.25$0.23 → $0.80+240%

What we record over time

We copy every provider's pricing page into a dated snapshot and never edit an old one. Snapshots so far: Oct 17, 2026, Oct 23, 2026. The difference between two snapshots is the changelog; each model page charts its own checks. Prices the providers have announced for later:

  • Dec 1, 2026: Gemini 4 Argon goes to $4 input / $20 output per 1M tokens (announced on the provider page; $2 / $10 until then). Source
  • Jan 1, 2027: Gemini 3.6 Flash goes to $1.5 input / $7.5 output per 1M tokens (announced on the provider page; $0.75 / $3.75 until then). Source
  • Jan 1, 2027: Gemini 3.8 Flash goes to $1.5 input / $7.5 output per 1M tokens (announced on the provider page; $0.75 / $3.75 until then). Source

Read this before you quote it

  • Today's price, not the launch price. Older models are shown at what the provider charges for them today. Some were cut after launch (GPT-4o's newer snapshot lists $2.5 input; the May 2024 one still lists $5).
  • Order is by version, not by date. Most model ids carry no release date, so families are ordered by version number.
  • Price only. A newer model can be cheaper per token and still cost more per task if it writes longer answers or thinks more. We don't score quality.
  • Method in full: how we collect AI prices.

Sources

Up next

The decision this helps with: Three picks with a monthly cost for your volume, from the same prices.LLM API prices, and the cheapest model for your jobLLM API pricing for 71 models per 1M tokens, checked at the source. Pick for me: your job, volume and budget give 3 models with the monthly cost. Filters, CSV.Read it next →

Also worth your time

Get the receipts by email.

What my AI agents tried this week, with the real numbers: what worked, what flopped, what made money. One email when there is something worth reading.

No spam, never sold. Every email has a one-click unsubscribe.

Cite this

Free to cite and reuse under CC BY 4.0 with a link back. How it was collected: methods. Spotted an error? Corrections.

Did this answer your question?

Popular in the shop

All 124 tools →
Every App$39AI Builder Kit$29Is Your Landing Page Cooked? 26-Point Roast Scorecard$5Blueprint$9Plainspoken$7Promptbook$6Envelope$9Paid$8Freelance Rate Calculator$5Homebrew Recipe Calculator$5Glaze Chemist$8Should I Deploy This? 18-Point Deploy Risk Checklist$5

Browse: Bundles and kitsHome, family and schoolMoney and businessAI toolsSports and fitnessMaking and hobbiesFood and drinkIndie dev and launchTrading and quant

Free: Landing page teardown iPhone app cost calculator LLM API cost calculator Paid online: $49 Launch Check App builds Articles FAQ