SaaS Pricing Tracker

AI model APIs pricing

Model APIs are priced per million tokens, with input and output charged separately and each model tier a fixed step apart. It is pure usage pricing: no seats and no plans, just a rate card that gets rewritten with every model generation, usually downwards for the same capability.

AI price index and 2030 predictionPrice history

AI price index

Where LLM API prices go next, to 2028 and beyond

$2

per million input tokens for the cheapest frontier LLM, 90% less than Nov 2023. Our view for 2028: $0.50 to $4, and $0.20 to $5 by 2030.

See the AI price index and prediction
$60$0.50–$4$0.20–$52021Today20282030

Exhibit 1

AI model APIs prices over time

The cheapest plan each product still sells, traced back through archived copies of its pricing page. Click a legend entry to hide a product.

Changed between readings
$0.05$0.10$0.20$0.50$120222023202420252026$0.20 → $0.10 · Sep 2026$1.38 Amazon Bedrock Amazon Titan and Nova, input +72%$1.25 xAI API grok-4.3, input$1 Claude API Haiku tier, input +25%$1 Perplexity Sonar API Sonar, input$0.75 Gemini API Gemini 3.8 Flash, input$0.75 Vertex AI Gemini Flash tier, input +50%$0.15 Mistral API Mistral Small 4, input$0.10 Azure OpenAI Cheapest model, input −50%$0.10 OpenAI API Cheapest model, input −50%$0.04 TypeSafe AI Jev
Note: Amazon Bedrock: USD per million tokens, on-demand in US East (N. Virginia), global cross-region rate where offered. Azure OpenAI: USD per million tokens, standard global deployment (regional and Data Zone before global existed). Claude API: USD per million tokens, standard processing. Gemini API: USD per million tokens, paid tier, prompts up to 200K tokens. Mistral API: USD per million tokens. OpenAI API: USD per million tokens, standard processing. Perplexity Sonar API: USD per million tokens, before per request fees. TypeSafe AI: USD per million input tokens; output tokens are free. Vertex AI: USD per million tokens, standard tier, prompts up to 200K tokens, global endpoint. xAI API: USD per million tokens, prompts under 200K tokens. Log scale.Source: live and archived pricing pages; SaaS Pricing Tracker analysis.
Show the data as a table
ObservedAmazon Bedrock Amazon Titan and Nova, inputAzure OpenAI Cheapest model, inputClaude API Haiku tier, inputGemini API Gemini 3.8 Flash, inputMistral API Mistral Small 4, inputOpenAI API Cheapest model, inputPerplexity Sonar API Sonar, inputTypeSafe AI JevVertex AI Gemini Flash tier, inputxAI API grok-4.3, input
Nov 20, 2021$0.80
Nov 3, 2022$0.40
Sep 28, 2023$1.30
Oct 20, 2023$1.50
Dec 21, 2023$0.80
Jan 5, 2024$1
May 2, 2024$0.50
May 19, 2024$0.50
May 29, 2024$0.20
Jun 20, 2024$0.25
Aug 1, 2024$0.15
Nov 12, 2024$0.15
Dec 3, 2024$1
Jan 9, 2025$0.80
Mar 2, 2025$0.15
Jun 3, 2025$0.80
Jun 6, 2025$0.15
Aug 1, 2025$0.05
Sep 2, 2025$0.30
Oct 1, 2025$0.80
Nov 1, 2025$0.05
Nov 11, 2025$0.05
Dec 23, 2025$1.38
Jan 1, 2026$1
Mar 1, 2026$0.50
Apr 1, 2026$1$0.50
Jul 1, 2026$1
Aug 1, 2026$0.20
Sep 1, 2026$0.10
Sep 21, 2026$1$0.75$0.15$0.20$1$0.04$1.25
Sep 30, 2026$1.38
Oct 3, 2026$0.10$1$0.75$0.15$0.10$1$0.04$0.75$1.25
Every price change on record

Exhibit 2

Azure OpenAI is the cheapest way in at $0.10, against $1 for xAI API

Entry price of each product, USD per million tokens

  • Azure OpenAIGPT-6 Luna$0.10
  • OpenAI APIgpt-6-luna$0.10
  • Mistral APIMistral Small 4$0.15
  • Gemini APIGemini 3.1 Flash-Lite$0.25
  • Amazon BedrockAmazon Nova 2 Lite$0.30
  • DeepSeek APIdeepseek-flash$0.30
  • Vertex AIGemini 3.8 Flash$0.75
  • Claude APIHaiku 4.5$1
  • Perplexity Sonar APISonar$1
  • xAI APIgrok-build-0.1$1

Exhibit 3

The priciest model is not the best: Opus 5.5 scores 58 at $8 per million tokens, Fable 5.1 costs $20 and scores 53

Artificial Analysis Intelligence Index (higher is better) against blended price, $ per million tokens at 3:1 input to output, log scale

Best score for the priceOther modelsAI Price Index scenario 2028: same scores, 25x cheaper
03060$0.01$0.1$1$102028 scenario2028: 25× cheaperBlended price per million tokensOpus 5.5: 58, $8 per million tokensOpus 5.5Sonnet 5.5: 56, $4 per million tokensSonnet 5.5Fable 5.1, gpt-6-astra: 53, $20 per million tokensFable 5.1, gpt-6-astraHaiku 4.5: 17, $2 per million tokensHaiku 4.5gpt-6.1-sol: 52, $4 per million tokensgpt-6.1-solgpt-5.6-terra: 42, $4.5 per million tokensgpt-5.6-terragpt-6-luna: 38, $0.2 per million tokensgpt-6-lunaGemini 3.8 Flash: 41, $1.5 per million tokensGemini 3.8 FlashGemini 3.1 Pro Preview: 30, $4.5 per million tokensGemini 3.1 Pro PreviewGemini 3.5 Flash-Lite: 22, $0.85 per million tokensGemini 3.5 Flash-Litegrok-4.7: 46, $3 per million tokensgrok-4.7deepseek-flash: 39, $0.52 per million tokensdeepseek-flashdeepseek-v4-pro: 36, $1.98 per million tokensdeepseek-v4-proMistral Medium 3.5: 14, $3 per million tokensMistral Medium 3.5Mistral Small 4: 11, $0.26 per million tokensMistral Small 4Mistral Large 3: 9, $0.75 per million tokensMistral Large 3Ministral 3 8B: 5, $0.15 per million tokensMinistral 3 8BNova 2 Pro: 14, $3.53 per million tokensNova 2 ProNova 2 Lite: 13, $0.85 per million tokensNova 2 Lite

Source: Artificial Analysis Intelligence Index, read Oct 3, 2026, highest reasoning setting per model; list prices from each pricing page; 2028 shift is the central scenario of the AI price index.

Exhibit 4

Resellers charge the model maker's list price to the cent, plus 10 percent to keep data in one region

Flagship model, $ per million input tokens, by where you buy it

  • OpenAI APIdirect$10.00
  • Azure OpenAIglobal deployment$10.00
  • Azure OpenAIData Zone (in-region)$11.00
  • Claude APIdirect$10.00
  • Amazon Bedrockglobal inference$10.00
  • Amazon Bedrockregional endpoint$11.00
  • Gemini APIdirect$2.00
  • Vertex AIglobal endpoint$2.00

Source: live and archived pricing pages of OpenAI, Azure, Anthropic, AWS and Google; regional and Data Zone rates are the published 10 percent premium.

AI model APIs compared

Current list prices as last read from each pricing page.

ProductStarts atTop listed planPriced byFree planHistory
Amazon Bedrock$0.30 per million input tokens, $2.5 per million outputAmazon Nova 2 Lite, USD per million tokens, on-demand in US East (N. Virginia), global cross-region rate where offered$10 per million input tokens, $50 per million output, globalClaude Fable 5.1Usage basedNosince 2023
Azure OpenAI$0.10 per million input tokens, $0.5 per million output, globalGPT-6 Luna, USD per million tokens, standard global deployment (regional and Data Zone before global existed)$10 per million input tokens, $50 per million output, globalGPT-6 AstraUsage basedNosince 2023
Claude API$1 per million input tokens, $5 per million outputHaiku 4.5, USD per million tokens, standard processing$10 per million input tokens, $50 per million outputFable 5.1Usage basedNosince 2024
Gemini API$0.25 per million input tokens, $1.5 per million outputGemini 3.1 Flash-Lite, USD per million tokens, paid tier, prompts up to 200K tokens$2 per million input tokens, $12 per million output for prompts up to 200K tokensGemini 3.1 Pro PreviewUsage basedYessince 2026
Mistral API$0.15 per million input tokens, $0.6 per million outputMistral Small 4, USD per million tokens$1.50 per million input tokens, $7.5 per million outputMistral Medium 3.5Usage basedNosince 2026
OpenAI API$0.10 per million input tokens, $0.5 per million outputgpt-6-luna, USD per million tokens, standard processing$10 per million input tokens, $50 per million outputgpt-6-astraUsage basedNosince 2021
Perplexity Sonar API$1 per million input tokens, $1 per million output, plus $5 to $12 per 1,000 requestsSonar, USD per million tokens, before per request fees$3 per million input tokens, $15 per million output, plus $6 to $14 per 1,000 requestsSonar ProUsage basedNosince 2026
TypeSafe AI$0.042 per million input tokens, output tokens freeJev, USD per million input tokens; output tokens are freecontact salesUsage basedNosince 2026
Vertex AI$0.75 per million input tokens, $3.75 per million output, introductory until 2027Gemini 3.8 Flash, USD per million tokens, standard tier, prompts up to 200K tokens, global endpoint$2 per million input tokens, $12 per million outputGemini 3.1 Pro PreviewUsage basedNosince 2025
xAI API$1 per million input tokens, $2 per million outputgrok-build-0.1, USD per million tokens, prompts under 200K tokens$2 per million input tokens, $6 per million outputgrok-4.6Usage basedNosince 2026
DeepSeek API$0.30 per million input tokens at peak, $1.20 per million output, half price off peakdeepseek-flash, USD per million tokens, peak rate, cache miss$1.32 per million input tokens at peak, $3.96 per million output, half price off peakdeepseek-v4-proUsage basedNoread Oct 2026

Head to head

Two products on one chart, plans side by side.

LLM API pricing history and trend

Every list price change on record for AI model APIs, with a trend line for each plan. Each change is read from an archived copy of the pricing page.

dated price changes
36
increases / cuts
39 / 46
earliest snapshot
2021
median change a year
+4%

Exhibit 5

2 of 5 products raised their headline price over the last three years

Compound yearly change of each product's headline plan, last three years

Price fellPrice rose
  • Vertex AIGemini Pro tier, input+43% a year
  • Amazon BedrockClaude flagship, input+8% a year
  • OpenAI APIFlagship model, input0% a year
  • Claude APITop model, input-16% a year
  • Azure OpenAIFlagship model, input-31% a year
Source: Live and archived pricing pages; SaaS Pricing Tracker analysis.

Exhibit 6

Trend, plan by plan

Compound yearly change from each plan’s first to latest price within the last three years.

Product and planFromNowPer year
Claude APIUSD per million tokens, standard processing
Top model, input$15Jun 2024$10-16%
Top model, output$75Jun 2024$50-16%
Sonnet tier, input$3Jun 2024$2-16%
Sonnet tier, output$15Jun 2024$10-16%
Haiku tier, input$0.25Jun 2024$1+83%
Haiku tier, output$1.25Jun 2024$5+83%
OpenAI APIUSD per million tokens, standard processing
Flagship model, input$10May 2024$100%
Flagship model, output$30May 2024$50+23%
Cheapest model, input$0.5May 2024$0.1-49%
Cheapest model, output$1.5May 2024$0.5-36%
Amazon BedrockUSD per million tokens, on-demand in US East (N. Virginia), global cross-region rate where offered
Claude flagship, input$8Dec 2023$10+8%
Claude flagship, output$24Dec 2023$50+30%
Amazon Titan and Nova, input$0.8Dec 2023$1.38+22%
Amazon Titan and Nova, output$1.6Dec 2023$10+94%
Vertex AIUSD per million tokens, standard tier, prompts up to 200K tokens, global endpoint
Gemini Pro tier, input$1.25Jun 2025$2+43%
Gemini Pro tier, output$10Jun 2025$12+15%
Gemini Flash tier, input$0.15Mar 2025$0.75+176%
Gemini Flash tier, output$0.6Mar 2025$3.75+217%
Azure OpenAIUSD per million tokens, standard global deployment (regional and Data Zone before global existed)
Flagship model, input$30Oct 2023$10-31%
Flagship model, output$60Oct 2023$50-6%
Cheapest model, input$1.5Oct 2023$0.1-60%
Cheapest model, output$2Oct 2023$0.5-37%

Exhibit 7

Amazon Bedrock Amazon Titan and Nova output rose 488% since 2023, from $1.70 to $10

Amazon Bedrock price history, USD per million tokens, on-demand in US East (N. Virginia), global cross-region rate where offered

Amazon Bedrock is where AWS customers buy Claude, Amazon Nova and other models without a separate vendor contract. Claude's top tier on Bedrock went from $8 and $24 per million tokens for Claude 2.1 to $15 and $75 for Opus 4, and now tracks Anthropic's own API exactly at $10 and $50 for Fable 5.1.

Changed between readings
$0.20$0.30$0.50$0.70$1$1.50$2$3$5$7$10$15$20$30$50$70202420252026$50 Claude flagship, output +100%$10 Claude flagship, input +100%$10 Amazon Titan and Nova, output +213%$1.38 Amazon Titan and Nova, input +72%
Note: USD per million tokens, on-demand in US East (N. Virginia), global cross-region rate where offered. Log scale.Source: Internet Archive snapshots of aws.amazon.com/bedrock/pricing and the AWS Price List API version history; SaaS Pricing Tracker analysis.
Show the data as a table
ObservedClaude flagship, inputClaude flagship, outputAmazon Titan and Nova, inputAmazon Titan and Nova, output
Sep 28, 2023$1.30$1.70
Sep 29, 2023$11.02$32.68
Dec 2, 2023$8$24
Dec 21, 2023$0.80$1.60
Mar 4, 2024$3$15
May 29, 2024$0.20$0.60
Sep 1, 2024$3$15
Jan 9, 2025$0.80$3.20
Jun 3, 2025$15$75
Nov 27, 2025$5$25
Dec 23, 2025$1.38$10
Sep 22, 2026$10$50
Sep 30, 2026$10$50$1.38$10

Claude flagship, input appeared by Sep 2023 at $11.02. It came down to $8 by Dec 2023 (-27%), then came down to $3 by Mar 2024 (-62%), then went up to $15 by Jun 2025 (+400%), then came down to $5 by Nov 2025 (-67%), and finally went up to $10 by Sep 2026 (+100%).

Claude flagship, output appeared by Sep 2023 at $32.68. It came down to $24 by Dec 2023 (-27%), then came down to $15 by Mar 2024 (-37%), then went up to $75 by Jun 2025 (+400%), then came down to $25 by Nov 2025 (-67%), and finally went up to $50 by Sep 2026 (+100%).

Amazon Titan and Nova, input has been listed since Sep 2023, starting at $1.30. It came down to $0.80 by Dec 2023 (-38%), then came down to $0.20 by May 2024 (-75%), then went up to $0.80 by Jan 2025 (+300%), and finally went up to $1.38 by Dec 2025 (+72%).

Amazon Titan and Nova, output has been listed since Sep 2023, starting at $1.70. It came down to $1.60 by Dec 2023 (-6%), then came down to $0.60 by May 2024 (-62%), then went up to $3.20 by Jan 2025 (+433%), and finally went up to $10 by Dec 2025 (+213%).

The Claude line follows the most capable Claude model sold in US East: Claude 2 at launch, Claude 2.1, Claude 3 Sonnet and 3.5 Sonnet while Claude 3 Opus was only offered in other regions, then Opus 4, Opus 4.5 and Fable 5.1. Since late 2025 Claude is billed through AWS Marketplace and its global rate matches Anthropic's own API, with regional endpoints 10 percent dearer. The Amazon line follows Amazon's own mid-range model: Titan Text Express, then Nova Pro from January 2025 and Nova 2 Pro from December 2025. AWS publishes a new price list version every few weeks, so dates are when a version first carried the price.

Full Amazon Bedrock pricing analysisVisit Amazon Bedrock

Exhibit 8

Azure OpenAI Cheapest model input fell 93% since 2023, from $1.50 to $0.10

Azure OpenAI price history, USD per million tokens, standard global deployment (regional and Data Zone before global existed)

Azure OpenAI is how Microsoft customers buy GPT models, with global prices that have tracked OpenAI's own API since 2024. The flagship fell from $30 per million input tokens for GPT-4 to $1.25 for GPT-5 and has climbed back to $10 with GPT-6 Astra.

Changed between readings
$0.10$1$10202420252026$1.20 → $0.50 · Sep 2026$50 Flagship model, output +67%$10 Flagship model, input +100%$0.50 Cheapest model, output −58%$0.10 Cheapest model, input −50%
Note: USD per million tokens, standard global deployment (regional and Data Zone before global existed). Log scale.Source: Internet Archive snapshots of the Azure OpenAI pricing page, the Azure Retail Prices API and the live page; SaaS Pricing Tracker analysis.
Show the data as a table
ObservedFlagship model, inputFlagship model, outputCheapest model, inputCheapest model, output
Oct 20, 2023$30$60$1.50$2
Jan 5, 2024$10$30$1$2
May 19, 2024$0.50$1.50
Aug 1, 2024$5$15$0.15$0.60
Nov 3, 2024$2.50$10
Aug 1, 2025$1.25$10$0.05$0.40
Nov 1, 2025$1.25$10$0.05$0.40
Dec 1, 2025$1.75$14
May 1, 2026$5$30
Aug 1, 2026$0.20$1.20
Sep 1, 2026$10$50$0.10$0.50
Oct 3, 2026$10$50$0.10$0.50

Flagship model, input has been listed since Oct 2023, starting at $30. It came down to $10 by Jan 2024 (-67%), then came down to $5 by Aug 2024 (-50%), then came down to $2.50 by Nov 2024 (-50%), then came down to $1.25 by Aug 2025 (-50%), then went up to $1.75 by Dec 2025 (+40%), then went up to $5 by May 2026 (+186%), and finally went up to $10 by Sep 2026 (+100%).

Flagship model, output has been listed since Oct 2023, starting at $60. It came down to $30 by Jan 2024 (-50%), then came down to $15 by Aug 2024 (-50%), then came down to $10 by Nov 2024 (-33%), then went up to $14 by Dec 2025 (+40%), then went up to $30 by May 2026 (+114%), and finally went up to $50 by Sep 2026 (+67%).

Cheapest model, input has been listed since Oct 2023, starting at $1.50. It came down to $1 by Jan 2024 (-33%), then came down to $0.50 by May 2024 (-50%), then came down to $0.15 by Aug 2024 (-70%), then came down to $0.05 by Aug 2025 (-67%), then went up to $0.20 by Aug 2026 (+300%), and finally came down to $0.10 by Sep 2026 (-50%).

Cheapest model, output has been listed since Oct 2023, starting at $2. It came down to $1.50 by May 2024 (-25%), then came down to $0.60 by Aug 2024 (-60%), then came down to $0.40 by Aug 2025 (-33%), then went up to $1.20 by Aug 2026 (+200%), and finally came down to $0.50 by Sep 2026 (-58%).

Azure quoted prices per 1,000 tokens until late 2024; they are converted to millions here. The flagship line follows GPT-4, GPT-4 Turbo, GPT-4o, GPT-5, GPT-5.2, GPT-5.5 and GPT-6 Astra; the cheapest line follows GPT-3.5 Turbo, GPT-4o mini, GPT-5 nano, GPT-5.6 Luna and GPT-6 Luna. Points from August 2025 onwards that are not archive snapshots use the effective start date of the meter in the Azure Retail Prices API. Global rates match OpenAI's own API, and Data Zone deployments cost 10 percent more.

Full Azure OpenAI pricing analysisVisit Azure OpenAI

Exhibit 9

Claude API Haiku tier input rose 300% since 2024, from $0.25 to $1

Claude API price history, USD per million tokens, standard processing

The Claude API is how developers pay for Anthropic's models: per million tokens, with input and output priced separately and every model tier a fixed step apart. Since Opus 4.5 in late 2025 the top tier has dropped from $15 to $5 per million input tokens while a new tier above it, Fable 5.1, took the $10 slot.

Changed between readings
$0.50$1$2$5$10$20$5020252026$50 Top model, output +100%$10 Top model, input +100%$10 Sonnet tier, output −33%$5 Haiku tier, output +25%$2 Sonnet tier, input −33%$1 Haiku tier, input +25%
Note: USD per million tokens, standard processing. Log scale.Source: Internet Archive snapshots of anthropic.com/pricing and claude.com/pricing, and the live page; SaaS Pricing Tracker analysis.
Show the data as a table
ObservedTop model, inputTop model, outputSonnet tier, inputSonnet tier, outputHaiku tier, inputHaiku tier, output
Jun 20, 2024$15$75$3$15$0.25$1.25
Dec 3, 2024$15$75$3$15$1$5
Jun 3, 2025$15$75$3$15$0.80$4
Oct 1, 2025$15$75$3$15$0.80$4
Jan 1, 2026$5$25$3$15$1$5
Apr 1, 2026$5$25$3$15$1$5
Jul 1, 2026$10$50$2$10$1$5
Sep 21, 2026$10$50$2$10$1$5
Oct 3, 2026$10$50$2$10$1$5

Top model, input has been listed since Jun 2024, starting at $15. It came down to $5 by Jan 2026 (-67%), and finally went up to $10 by Jul 2026 (+100%).

Top model, output has been listed since Jun 2024, starting at $75. It came down to $25 by Jan 2026 (-67%), and finally went up to $50 by Jul 2026 (+100%).

Sonnet tier, input has been listed since Jun 2024, starting at $3. It came down to $2 by Jul 2026 (-33%).

Sonnet tier, output has been listed since Jun 2024, starting at $15. It came down to $10 by Jul 2026 (-33%).

Haiku tier, input has been listed since Jun 2024, starting at $0.25. It went up to $1 by Dec 2024 (+300%), then came down to $0.80 by Jun 2025 (-20%), and finally went up to $1 by Jan 2026 (+25%).

Haiku tier, output has been listed since Jun 2024, starting at $1.25. It went up to $5 by Dec 2024 (+300%), then came down to $4 by Jun 2025 (-20%), and finally went up to $5 by Jan 2026 (+25%).

The top line follows the highest priced current model: Opus 3, 4 and 4.1 at $15 and $75, Opus 4.5 and 4.6 at a third of that from late 2025, then Fable 5 and 5.1 at $10 and $50 from mid 2026. The Sonnet tier held $3 and $15 for two years until Sonnet 5 launched at $2 and $10 in 2026, first as an introductory price. The Haiku tier went from $0.25 and $1.25 to $1 and $5 with Haiku 3.5, was cut to $0.80 and $4 in 2025 and returned to $1 and $5 with Haiku 4.5. The API rate card moved from anthropic.com to claude.com in autumn 2025.

Full Claude API pricing analysisVisit Claude API

Exhibit 10

Gemini API has kept Gemini 3.1 Pro Preview input at $2 since 2026

Gemini API price history, USD per million tokens, paid tier, prompts up to 200K tokens

The Gemini API is Google's per token offer to developers and the only one of the big three with a free tier on most models. Pro models charge more above 200,000 tokens of prompt, and the newest Flash models are sold at an introductory rate that Google has already announced will double in 2027.

$1$2$5$10Nov 2026$12 Gemini 3.1 Pro Preview, output$3.75 Gemini 3.8 Flash, output$2 Gemini 3.1 Pro Preview, input$0.75 Gemini 3.8 Flash, input
Note: USD per million tokens, paid tier, prompts up to 200K tokens. Log scale.Source: the live Gemini API pricing page; SaaS Pricing Tracker analysis.
Show the data as a table
ObservedGemini 3.1 Pro Preview, inputGemini 3.1 Pro Preview, outputGemini 3.8 Flash, inputGemini 3.8 Flash, output
Sep 21, 2026$2$12$0.75$3.75
Oct 3, 2026$2$12$0.75$3.75

Gemini 3.1 Pro Preview, input has been listed since Sep 2026, starting at $2. The price was unchanged through Oct 2026.

Gemini 3.1 Pro Preview, output has been listed since Sep 2026, starting at $12. The price was unchanged through Oct 2026.

Gemini 3.8 Flash, input has been listed since Sep 2026, starting at $0.75. The price was unchanged through Oct 2026.

Gemini 3.8 Flash, output has been listed since Sep 2026, starting at $3.75. The price was unchanged through Oct 2026.

First read on 21 September 2026, so the chart starts now.

Full Gemini API pricing analysisVisit Gemini API

Exhibit 11

Mistral API has kept Mistral Large 3 input at $0.50 since 2026

Mistral API price history, USD per million tokens

Mistral's La Plateforme prices its models far below the US labs per million tokens, and unusually its flagship Large 3 costs less than the mid tier Medium 3.5. Open weight models like Ministral sit at the bottom with the same price for input and output.

$0.20$0.50$1$2$5Nov 2026$7.50 Mistral Medium 3.5, output$1.50 Mistral Large 3, output$1.50 Mistral Medium 3.5, input$0.60 Mistral Small 4, output$0.50 Mistral Large 3, input$0.15 Mistral Small 4, input
Note: USD per million tokens. Log scale.Source: the live Mistral pricing documentation; SaaS Pricing Tracker analysis.
Show the data as a table
ObservedMistral Large 3, inputMistral Large 3, outputMistral Medium 3.5, inputMistral Medium 3.5, outputMistral Small 4, inputMistral Small 4, output
Sep 21, 2026$0.50$1.50$1.50$7.50$0.15$0.60
Oct 3, 2026$0.50$1.50$1.50$7.50$0.15$0.60

Mistral Large 3, input has been listed since Sep 2026, starting at $0.50. The price was unchanged through Oct 2026.

Mistral Large 3, output has been listed since Sep 2026, starting at $1.50. The price was unchanged through Oct 2026.

Mistral Medium 3.5, input has been listed since Sep 2026, starting at $1.50. The price was unchanged through Oct 2026.

Mistral Medium 3.5, output has been listed since Sep 2026, starting at $7.50. The price was unchanged through Oct 2026.

Mistral Small 4, input has been listed since Sep 2026, starting at $0.15. The price was unchanged through Oct 2026.

Mistral Small 4, output has been listed since Sep 2026, starting at $0.60. The price was unchanged through Oct 2026.

First read on 21 September 2026, so the chart starts now.

Full Mistral API pricing analysisVisit Mistral API

Exhibit 12

OpenAI API Cheapest model input fell 88% since 2021, from $0.80 to $0.10

OpenAI API price history, USD per million tokens, standard processing

The OpenAI API is the developer side of ChatGPT's maker, priced per million tokens with cached input at a tenth of the standard input rate. Batch and Flex processing cost half, Fast mode costs double, and the flagship has been repriced with almost every model generation since GPT-3 charged per thousand tokens.

Changed between readings
$0.10$1$1020222023202420252026$1.20 → $0.50 · Oct 2026$50 Flagship model, output +67%$10 Flagship model, input +100%$0.50 Cheapest model, output −58%$0.10 Cheapest model, input −50%
Note: USD per million tokens, standard processing. Log scale.Source: Internet Archive snapshots of openai.com/api/pricing and the live platform pricing page; SaaS Pricing Tracker analysis.
Show the data as a table
ObservedFlagship model, inputFlagship model, outputCheapest model, inputCheapest model, output
Nov 20, 2021$60$60$0.80$0.80
Nov 3, 2022$20$20$0.40$0.40
May 2, 2024$10$30$0.50$1.50
Nov 12, 2024$2.50$10$0.15$0.60
Nov 11, 2025$1.25$10$0.05$0.40
May 9, 2026$5$30
Sep 21, 2026$10$50$0.20$1.20
Oct 3, 2026$10$50$0.10$0.50

Flagship model, input has been listed since Nov 2021, starting at $60. It came down to $20 by Nov 2022 (-67%), then came down to $10 by May 2024 (-50%), then came down to $2.50 by Nov 2024 (-75%), then came down to $1.25 by Nov 2025 (-50%), then went up to $5 by May 2026 (+300%), and finally went up to $10 by Sep 2026 (+100%).

Flagship model, output has been listed since Nov 2021, starting at $60. It came down to $20 by Nov 2022 (-67%), then went up to $30 by May 2024 (+50%), then came down to $10 by Nov 2024 (-67%), then went up to $30 by May 2026 (+200%), and finally went up to $50 by Sep 2026 (+67%).

Cheapest model, input has been listed since Nov 2021, starting at $0.80. It came down to $0.40 by Nov 2022 (-50%), then went up to $0.50 by May 2024 (+25%), then came down to $0.15 by Nov 2024 (-70%), then came down to $0.05 by Nov 2025 (-67%), then went up to $0.20 by Sep 2026 (+300%), and finally came down to $0.10 by Oct 2026 (-50%).

Cheapest model, output has been listed since Nov 2021, starting at $0.80. It came down to $0.40 by Nov 2022 (-50%), then went up to $1.50 by May 2024 (+275%), then came down to $0.60 by Nov 2024 (-60%), then came down to $0.40 by Nov 2025 (-33%), then went up to $1.20 by Sep 2026 (+200%), and finally came down to $0.50 by Oct 2026 (-58%).

Early GPT-3 prices were quoted per thousand tokens with one rate for input and output; they are converted to millions here. The flagship line follows whichever model OpenAI listed first: Davinci, then GPT-4 Turbo, GPT-4o, GPT-5, the top GPT-5 series model in May 2026 and gpt-6-astra now. The cheapest line follows Ada, GPT-3.5 Turbo, GPT-4o mini, GPT-5 nano, gpt-5.6-luna and, from October 2026, gpt-6-luna. Flagship prices fell for four years and have risen since 2026.

Full OpenAI API pricing analysisVisit OpenAI API

Exhibit 13

Perplexity Sonar API has kept Sonar input at $1 since 2026

Perplexity Sonar API price history, USD per million tokens, before per request fees

Perplexity's Sonar API is a web-grounded answer engine for developers. Token prices look like everyone else's, but each request also carries a fee that depends on how much web context is pulled in, which makes it a hybrid of per token and per call pricing.

$1$2$5$10Nov 2026$15 Sonar Pro, output$3 Sonar Pro, input$1 Sonar, input$1 Sonar, output
Note: USD per million tokens, before per request fees. Log scale.Source: the live Sonar pricing documentation; SaaS Pricing Tracker analysis.
Show the data as a table
ObservedSonar, inputSonar, outputSonar Pro, inputSonar Pro, output
Sep 21, 2026$1$1$3$15
Oct 3, 2026$1$1$3$15

Sonar, input has been listed since Sep 2026, starting at $1. The price was unchanged through Oct 2026.

Sonar, output has been listed since Sep 2026, starting at $1. The price was unchanged through Oct 2026.

Sonar Pro, input has been listed since Sep 2026, starting at $3. The price was unchanged through Oct 2026.

Sonar Pro, output has been listed since Sep 2026, starting at $15. The price was unchanged through Oct 2026.

First read on 21 September 2026, so the chart starts now.

Full Perplexity Sonar API pricing analysisVisit Perplexity Sonar API

Exhibit 14

TypeSafe AI has kept Jev at $0.04 since 2026

TypeSafe AI price history, USD per million input tokens; output tokens are free

TypeSafe AI sells Jev, a model that answers typed questions about program state with structured, calibrated decisions rather than free text, aimed at automation inside software. There are no plans or seats: usage is metered at $42 per billion input tokens and output is not charged, which the company positions against the per-token prices of general purpose language models.

Jev
$0$0.25$0.50$0.75$1$1.25Nov 2026$0.04 Jev
Note: USD per million input tokens; output tokens are free.Source: the live typesafe.ai homepage, which carries the only published price; SaaS Pricing Tracker analysis.
Show the data as a table
ObservedJev
Sep 21, 2026$0.04
Oct 3, 2026$0.04

The plan has been listed since Sep 2026, starting at $0.04. The price was unchanged through Oct 2026.

Jev launched in early access on 15 September 2026 at $0.042 per million input tokens ($42 per billion) with no charge for output, and the company says it expects the price to go down rather than up. There is no pricing page yet, so the chart starts at the launch price.

Full TypeSafe AI pricing analysisVisit TypeSafe AI

Exhibit 15

Vertex AI Gemini Flash tier output rose 525% since 2025, from $0.60 to $3.75

Vertex AI price history, USD per million tokens, standard tier, prompts up to 200K tokens, global endpoint

Vertex AI is the Google Cloud route to Gemini, priced the same as Google's developer API but billed through a Cloud account. Gemini Pro went from $1.25 to $2 per million input tokens with Gemini 3, and the Flash tier has climbed from $0.15 to $0.75.

Changed between readings
$0.20$0.50$1$2$5$10Apr 2025JulOctJan 2026AprJulOct$12 Gemini Pro tier, output +20%$3.75 Gemini Flash tier, output +25%$2 Gemini Pro tier, input +60%$0.75 Gemini Flash tier, input +50%
Note: USD per million tokens, standard tier, prompts up to 200K tokens, global endpoint. Log scale.Source: Internet Archive snapshots of cloud.google.com/vertex-ai/generative-ai/pricing and the live page; SaaS Pricing Tracker analysis.
Show the data as a table
ObservedGemini Pro tier, inputGemini Pro tier, outputGemini Flash tier, inputGemini Flash tier, output
Mar 2, 2025$0.15$0.60
Jun 6, 2025$1.25$10$0.15$0.60
Sep 2, 2025$1.25$10$0.30$2.50
Dec 7, 2025$2$12
Mar 1, 2026$2$12$0.50$3
Apr 1, 2026$2$12$0.50$3
Oct 3, 2026$2$12$0.75$3.75

Gemini Pro tier, input appeared by Jun 2025 at $1.25. It went up to $2 by Dec 2025 (+60%).

Gemini Pro tier, output appeared by Jun 2025 at $10. It went up to $12 by Dec 2025 (+20%).

Gemini Flash tier, input has been listed since Mar 2025, starting at $0.15. It went up to $0.30 by Sep 2025 (+100%), then went up to $0.50 by Mar 2026 (+67%), and finally went up to $0.75 by Oct 2026 (+50%).

Gemini Flash tier, output has been listed since Mar 2025, starting at $0.60. It went up to $2.50 by Sep 2025 (+317%), then went up to $3 by Mar 2026 (+20%), and finally went up to $3.75 by Oct 2026 (+25%).

Vertex AI priced Gemini 1.0 and 1.5 per 1,000 characters rather than per token until 2025, so the chart starts with token pricing. The Pro line follows Gemini 2.5 Pro, 3 Pro Preview and 3.1 Pro Preview. The Flash line follows Gemini 2.0 Flash, 2.5 Flash (its non-thinking output rate in mid 2025), 3 Flash Preview and 3.8 Flash, whose $0.75 and $3.75 are introductory and rise to $1.50 and $7.50 on 1 January 2027. Prices match the Gemini Developer API.

Full Vertex AI pricing analysisVisit Vertex AI

Exhibit 16

xAI API has kept grok-4.6 input at $2 since 2026

xAI API price history, USD per million tokens, prompts under 200K tokens

The xAI API sells Grok models per million tokens with the same input and output split as its rivals, cached input at a discount and a doubled rate for long prompts. Its flagship sits between OpenAI's mid and top tiers on price.

$0$2$4$6$8Nov 2026$6 grok-4.6, output$6 grok-4.7, output$2.50 grok-4.3, output$2 grok-4.6, input$2 grok-4.7, input$1.25 grok-4.3, input
Note: USD per million tokens, prompts under 200K tokens.Source: the live xAI models page; SaaS Pricing Tracker analysis.
Show the data as a table
Observedgrok-4.6, inputgrok-4.6, outputgrok-4.3, inputgrok-4.3, outputgrok-4.7, inputgrok-4.7, output
Sep 21, 2026$2$6$1.25$2.50
Oct 3, 2026$2$6$1.25$2.50$2$6

grok-4.6, input has been listed since Sep 2026, starting at $2. The price was unchanged through Oct 2026.

grok-4.6, output has been listed since Sep 2026, starting at $6. The price was unchanged through Oct 2026.

grok-4.3, input has been listed since Sep 2026, starting at $1.25. The price was unchanged through Oct 2026.

grok-4.3, output has been listed since Sep 2026, starting at $2.50. The price was unchanged through Oct 2026.

grok-4.7, input appeared by Oct 2026 at $2.

grok-4.7, output appeared by Oct 2026 at $6.

First read on 21 September 2026, so the chart starts now.

Full xAI API pricing analysisVisit xAI API

2026

  1. OpenAI API cut the Cheapest model input and output prices

    • Cheapest model, input went from $0.20 to $0.10 (-50%)
    • Cheapest model, output went from $1.20 to $0.50 (-58%)

    USD per million tokens, standard processing. Archived pricing page

  2. Vertex AI raised the Gemini Flash tier input and output prices

    • Gemini Flash tier, input went from $0.50 to $0.75 (+50%)
    • Gemini Flash tier, output went from $3 to $3.75 (+25%)

    USD per million tokens, standard tier, prompts up to 200K tokens, global endpoint. Archived pricing page

  3. xAI API added the grok-4.7 input and output plans

    • grok-4.7, input listed at $2
    • grok-4.7, output listed at $6

    USD per million tokens, prompts under 200K tokens. Archived pricing page

  4. Amazon Bedrock raised the Claude flagship input and output prices

    • Claude flagship, input went from $5 to $10 (+100%)
    • Claude flagship, output went from $25 to $50 (+100%)

    USD per million tokens, on-demand in US East (N. Virginia), global cross-region rate where offered. Archived pricing page

  5. OpenAI API raised the Flagship model input and output and Cheapest model input and output prices

    • Flagship model, input went from $5 to $10 (+100%)
    • Flagship model, output went from $30 to $50 (+67%)
    • Cheapest model, input went from $0.05 to $0.20 (+300%)
    • Cheapest model, output went from $0.40 to $1.20 (+200%)

    USD per million tokens, standard processing. Archived pricing page

  6. Azure OpenAI changed its plans and prices

    • Flagship model, input went from $5 to $10 (+100%)
    • Flagship model, output went from $30 to $50 (+67%)
    • Cheapest model, input went from $0.20 to $0.10 (-50%)
    • Cheapest model, output went from $1.20 to $0.50 (-58%)

    USD per million tokens, standard global deployment (regional and Data Zone before global existed). Archived pricing page

  7. Azure OpenAI raised the Cheapest model input and output prices

    • Cheapest model, input went from $0.05 to $0.20 (+300%)
    • Cheapest model, output went from $0.40 to $1.20 (+200%)

    USD per million tokens, standard global deployment (regional and Data Zone before global existed). Archived pricing page

  8. Claude API changed its plans and prices

    • Top model, input went from $5 to $10 (+100%)
    • Top model, output went from $25 to $50 (+100%)
    • Sonnet tier, input went from $3 to $2 (-33%)
    • Sonnet tier, output went from $15 to $10 (-33%)

    USD per million tokens, standard processing. Archived pricing page

  9. OpenAI API raised the Flagship model input and output prices

    • Flagship model, input went from $1.25 to $5 (+300%)
    • Flagship model, output went from $10 to $30 (+200%)

    USD per million tokens, standard processing. Archived pricing page

  10. Azure OpenAI raised the Flagship model input and output prices

    • Flagship model, input went from $1.75 to $5 (+186%)
    • Flagship model, output went from $14 to $30 (+114%)

    USD per million tokens, standard global deployment (regional and Data Zone before global existed). Archived pricing page

  11. Vertex AI raised the Gemini Flash tier input and output prices

    • Gemini Flash tier, input went from $0.30 to $0.50 (+67%)
    • Gemini Flash tier, output went from $2.50 to $3 (+20%)

    USD per million tokens, standard tier, prompts up to 200K tokens, global endpoint. Archived pricing page

  12. Claude API changed its plans and prices

    • Top model, input went from $15 to $5 (-67%)
    • Top model, output went from $75 to $25 (-67%)
    • Haiku tier, input went from $0.80 to $1 (+25%)
    • Haiku tier, output went from $4 to $5 (+25%)

    USD per million tokens, standard processing. Archived pricing page

2025

  1. Amazon Bedrock raised the Amazon Titan and Nova input and output prices

    • Amazon Titan and Nova, input went from $0.80 to $1.38 (+72%)
    • Amazon Titan and Nova, output went from $3.20 to $10 (+213%)

    USD per million tokens, on-demand in US East (N. Virginia), global cross-region rate where offered. Archived pricing page

  2. Vertex AI raised the Gemini Pro tier input and output prices

    • Gemini Pro tier, input went from $1.25 to $2 (+60%)
    • Gemini Pro tier, output went from $10 to $12 (+20%)

    USD per million tokens, standard tier, prompts up to 200K tokens, global endpoint. Archived pricing page

  3. Azure OpenAI raised the Flagship model input and output prices

    • Flagship model, input went from $1.25 to $1.75 (+40%)
    • Flagship model, output went from $10 to $14 (+40%)

    USD per million tokens, standard global deployment (regional and Data Zone before global existed). Archived pricing page

  4. Amazon Bedrock cut the Claude flagship input and output prices

    • Claude flagship, input went from $15 to $5 (-67%)
    • Claude flagship, output went from $75 to $25 (-67%)

    USD per million tokens, on-demand in US East (N. Virginia), global cross-region rate where offered. Archived pricing page

  5. OpenAI API cut the Flagship model input and Cheapest model input and output prices

    • Flagship model, input went from $2.50 to $1.25 (-50%)
    • Cheapest model, input went from $0.15 to $0.05 (-67%)
    • Cheapest model, output went from $0.60 to $0.40 (-33%)

    USD per million tokens, standard processing. Archived pricing page

  6. Vertex AI raised the Gemini Flash tier input and output prices

    • Gemini Flash tier, input went from $0.15 to $0.30 (+100%)
    • Gemini Flash tier, output went from $0.60 to $2.50 (+317%)

    USD per million tokens, standard tier, prompts up to 200K tokens, global endpoint. Archived pricing page

  7. Azure OpenAI cut the Flagship model input and Cheapest model input and output prices

    • Flagship model, input went from $2.50 to $1.25 (-50%)
    • Cheapest model, input went from $0.15 to $0.05 (-67%)
    • Cheapest model, output went from $0.60 to $0.40 (-33%)

    USD per million tokens, standard global deployment (regional and Data Zone before global existed). Archived pricing page

  8. Vertex AI added the Gemini Pro tier input and output plans

    • Gemini Pro tier, input listed at $1.25
    • Gemini Pro tier, output listed at $10

    USD per million tokens, standard tier, prompts up to 200K tokens, global endpoint. Archived pricing page

  9. Amazon Bedrock raised the Claude flagship input and output prices

    • Claude flagship, input went from $3 to $15 (+400%)
    • Claude flagship, output went from $15 to $75 (+400%)

    USD per million tokens, on-demand in US East (N. Virginia), global cross-region rate where offered. Archived pricing page

  10. Claude API cut the Haiku tier input and output prices

    • Haiku tier, input went from $1 to $0.80 (-20%)
    • Haiku tier, output went from $5 to $4 (-20%)

    USD per million tokens, standard processing. Archived pricing page

  11. Amazon Bedrock raised the Amazon Titan and Nova input and output prices

    • Amazon Titan and Nova, input went from $0.20 to $0.80 (+300%)
    • Amazon Titan and Nova, output went from $0.60 to $3.20 (+433%)

    USD per million tokens, on-demand in US East (N. Virginia), global cross-region rate where offered. Archived pricing page

2024

  1. Claude API raised the Haiku tier input and output prices

    • Haiku tier, input went from $0.25 to $1 (+300%)
    • Haiku tier, output went from $1.25 to $5 (+300%)

    USD per million tokens, standard processing. Archived pricing page

  2. OpenAI API cut the Flagship model input and output and Cheapest model input and output prices

    • Flagship model, input went from $10 to $2.50 (-75%)
    • Flagship model, output went from $30 to $10 (-67%)
    • Cheapest model, input went from $0.50 to $0.15 (-70%)
    • Cheapest model, output went from $1.50 to $0.60 (-60%)

    USD per million tokens, standard processing. Archived pricing page

  3. Azure OpenAI cut the Flagship model input and output prices

    • Flagship model, input went from $5 to $2.50 (-50%)
    • Flagship model, output went from $15 to $10 (-33%)

    USD per million tokens, standard global deployment (regional and Data Zone before global existed). Archived pricing page

  4. Azure OpenAI cut the Flagship model input and output and Cheapest model input and output prices

    • Flagship model, input went from $10 to $5 (-50%)
    • Flagship model, output went from $30 to $15 (-50%)
    • Cheapest model, input went from $0.50 to $0.15 (-70%)
    • Cheapest model, output went from $1.50 to $0.60 (-60%)

    USD per million tokens, standard global deployment (regional and Data Zone before global existed). Archived pricing page

  5. Amazon Bedrock cut the Amazon Titan and Nova input and output prices

    • Amazon Titan and Nova, input went from $0.80 to $0.20 (-75%)
    • Amazon Titan and Nova, output went from $1.60 to $0.60 (-62%)

    USD per million tokens, on-demand in US East (N. Virginia), global cross-region rate where offered. Archived pricing page

  6. Azure OpenAI cut the Cheapest model input and output prices

    • Cheapest model, input went from $1 to $0.50 (-50%)
    • Cheapest model, output went from $2 to $1.50 (-25%)

    USD per million tokens, standard global deployment (regional and Data Zone before global existed). Archived pricing page

  7. OpenAI API changed its plans and prices

    • Flagship model, input went from $20 to $10 (-50%)
    • Flagship model, output went from $20 to $30 (+50%)
    • Cheapest model, input went from $0.40 to $0.50 (+25%)
    • Cheapest model, output went from $0.40 to $1.50 (+275%)

    USD per million tokens, standard processing. Archived pricing page

  8. Amazon Bedrock cut the Claude flagship input and output prices

    • Claude flagship, input went from $8 to $3 (-62%)
    • Claude flagship, output went from $24 to $15 (-37%)

    USD per million tokens, on-demand in US East (N. Virginia), global cross-region rate where offered. Archived pricing page

  9. Azure OpenAI cut the Flagship model input and output and Cheapest model input prices

    • Flagship model, input went from $30 to $10 (-67%)
    • Flagship model, output went from $60 to $30 (-50%)
    • Cheapest model, input went from $1.50 to $1 (-33%)

    USD per million tokens, standard global deployment (regional and Data Zone before global existed). Archived pricing page

2023

  1. Amazon Bedrock cut the Amazon Titan and Nova input and output prices

    • Amazon Titan and Nova, input went from $1.30 to $0.80 (-38%)
    • Amazon Titan and Nova, output went from $1.70 to $1.60 (-6%)

    USD per million tokens, on-demand in US East (N. Virginia), global cross-region rate where offered. Archived pricing page

  2. Amazon Bedrock cut the Claude flagship input and output prices

    • Claude flagship, input went from $11.02 to $8 (-27%)
    • Claude flagship, output went from $32.68 to $24 (-27%)

    USD per million tokens, on-demand in US East (N. Virginia), global cross-region rate where offered. Archived pricing page

  3. Amazon Bedrock added the Claude flagship input and output plans

    • Claude flagship, input listed at $11.02
    • Claude flagship, output listed at $32.68

    USD per million tokens, on-demand in US East (N. Virginia), global cross-region rate where offered. Archived pricing page

2022

  1. OpenAI API cut the Flagship model input and output and Cheapest model input and output prices

    • Flagship model, input went from $60 to $20 (-67%)
    • Flagship model, output went from $60 to $20 (-67%)
    • Cheapest model, input went from $0.80 to $0.40 (-50%)
    • Cheapest model, output went from $0.80 to $0.40 (-50%)

    USD per million tokens, standard processing. Archived pricing page

AI model APIs pricing questions

What does a AI model API cost?
Paid plans start at $0.30 per million input tokens, $2.5 per million output for Amazon Bedrock (USD per million tokens, on-demand in US East (N. Virginia), global cross-region rate where offered), $0.10 per million input tokens, $0.5 per million output, global for Azure OpenAI (USD per million tokens, standard global deployment (regional and Data Zone before global existed)), $1 per million input tokens, $5 per million output for Claude API (USD per million tokens, standard processing), $0.25 per million input tokens, $1.5 per million output for Gemini API (USD per million tokens, paid tier, prompts up to 200K tokens), $0.15 per million input tokens, $0.6 per million output for Mistral API (USD per million tokens), $0.10 per million input tokens, $0.5 per million output for OpenAI API (USD per million tokens, standard processing), $1 per million input tokens, $1 per million output, plus $5 to $12 per 1,000 requests for Perplexity Sonar API (USD per million tokens, before per request fees), $0.042 per million input tokens, output tokens free for TypeSafe AI (USD per million input tokens; output tokens are free), $0.75 per million input tokens, $3.75 per million output, introductory until 2027 for Vertex AI (USD per million tokens, standard tier, prompts up to 200K tokens, global endpoint), $1 per million input tokens, $2 per million output for xAI API (USD per million tokens, prompts under 200K tokens), $0.30 per million input tokens at peak, $1.20 per million output, half price off peak for DeepSeek API (USD per million tokens, peak rate, cache miss). Prices are as shown on each pricing page when it was last read.
What is AI model API pricing based on?
Amazon Bedrock and Azure OpenAI and Claude API and Gemini API and Mistral API and OpenAI API and Perplexity Sonar API and TypeSafe AI and Vertex AI and xAI API and DeepSeek API: usage based.
Which ai model apis have a free plan?
Gemini API lists a free plan next to the paid tiers.
Which ai model apis have changed their prices?
Amazon Bedrock: Claude flagship, input appeared by Sep 2023 at $11.02. It came down to $8 by Dec 2023 (-27%), then came down to $3 by Mar 2024 (-62%), then went up to $15 by Jun 2025 (+400%), then came down to $5 by Nov 2025 (-67%), and finally went up to $10 by Sep 2026 (+100%). Claude flagship, output appeared by Sep 2023 at $32.68. It came down to $24 by Dec 2023 (-27%), then came down to $15 by Mar 2024 (-37%), then went up to $75 by Jun 2025 (+400%), then came down to $25 by Nov 2025 (-67%), and finally went up to $50 by Sep 2026 (+100%). Amazon Titan and Nova, input has been listed since Sep 2023, starting at $1.30. It came down to $0.80 by Dec 2023 (-38%), then came down to $0.20 by May 2024 (-75%), then went up to $0.80 by Jan 2025 (+300%), and finally went up to $1.38 by Dec 2025 (+72%). Amazon Titan and Nova, output has been listed since Sep 2023, starting at $1.70. It came down to $1.60 by Dec 2023 (-6%), then came down to $0.60 by May 2024 (-62%), then went up to $3.20 by Jan 2025 (+433%), and finally went up to $10 by Dec 2025 (+213%). Azure OpenAI: Flagship model, input has been listed since Oct 2023, starting at $30. It came down to $10 by Jan 2024 (-67%), then came down to $5 by Aug 2024 (-50%), then came down to $2.50 by Nov 2024 (-50%), then came down to $1.25 by Aug 2025 (-50%), then went up to $1.75 by Dec 2025 (+40%), then went up to $5 by May 2026 (+186%), and finally went up to $10 by Sep 2026 (+100%). Flagship model, output has been listed since Oct 2023, starting at $60. It came down to $30 by Jan 2024 (-50%), then came down to $15 by Aug 2024 (-50%), then came down to $10 by Nov 2024 (-33%), then went up to $14 by Dec 2025 (+40%), then went up to $30 by May 2026 (+114%), and finally went up to $50 by Sep 2026 (+67%). Cheapest model, input has been listed since Oct 2023, starting at $1.50. It came down to $1 by Jan 2024 (-33%), then came down to $0.50 by May 2024 (-50%), then came down to $0.15 by Aug 2024 (-70%), then came down to $0.05 by Aug 2025 (-67%), then went up to $0.20 by Aug 2026 (+300%), and finally came down to $0.10 by Sep 2026 (-50%). Cheapest model, output has been listed since Oct 2023, starting at $2. It came down to $1.50 by May 2024 (-25%), then came down to $0.60 by Aug 2024 (-60%), then came down to $0.40 by Aug 2025 (-33%), then went up to $1.20 by Aug 2026 (+200%), and finally came down to $0.50 by Sep 2026 (-58%). Claude API: Top model, input has been listed since Jun 2024, starting at $15. It came down to $5 by Jan 2026 (-67%), and finally went up to $10 by Jul 2026 (+100%). Top model, output has been listed since Jun 2024, starting at $75. It came down to $25 by Jan 2026 (-67%), and finally went up to $50 by Jul 2026 (+100%). Sonnet tier, input has been listed since Jun 2024, starting at $3. It came down to $2 by Jul 2026 (-33%). Sonnet tier, output has been listed since Jun 2024, starting at $15. It came down to $10 by Jul 2026 (-33%). Haiku tier, input has been listed since Jun 2024, starting at $0.25. It went up to $1 by Dec 2024 (+300%), then came down to $0.80 by Jun 2025 (-20%), and finally went up to $1 by Jan 2026 (+25%). Haiku tier, output has been listed since Jun 2024, starting at $1.25. It went up to $5 by Dec 2024 (+300%), then came down to $4 by Jun 2025 (-20%), and finally went up to $5 by Jan 2026 (+25%). OpenAI API: Flagship model, input has been listed since Nov 2021, starting at $60. It came down to $20 by Nov 2022 (-67%), then came down to $10 by May 2024 (-50%), then came down to $2.50 by Nov 2024 (-75%), then came down to $1.25 by Nov 2025 (-50%), then went up to $5 by May 2026 (+300%), and finally went up to $10 by Sep 2026 (+100%). Flagship model, output has been listed since Nov 2021, starting at $60. It came down to $20 by Nov 2022 (-67%), then went up to $30 by May 2024 (+50%), then came down to $10 by Nov 2024 (-67%), then went up to $30 by May 2026 (+200%), and finally went up to $50 by Sep 2026 (+67%). Cheapest model, input has been listed since Nov 2021, starting at $0.80. It came down to $0.40 by Nov 2022 (-50%), then went up to $0.50 by May 2024 (+25%), then came down to $0.15 by Nov 2024 (-70%), then came down to $0.05 by Nov 2025 (-67%), then went up to $0.20 by Sep 2026 (+300%), and finally came down to $0.10 by Oct 2026 (-50%). Cheapest model, output has been listed since Nov 2021, starting at $0.80. It came down to $0.40 by Nov 2022 (-50%), then went up to $1.50 by May 2024 (+275%), then came down to $0.60 by Nov 2024 (-60%), then came down to $0.40 by Nov 2025 (-33%), then went up to $1.20 by Sep 2026 (+200%), and finally came down to $0.50 by Oct 2026 (-58%). Vertex AI: Gemini Pro tier, input appeared by Jun 2025 at $1.25. It went up to $2 by Dec 2025 (+60%). Gemini Pro tier, output appeared by Jun 2025 at $10. It went up to $12 by Dec 2025 (+20%). Gemini Flash tier, input has been listed since Mar 2025, starting at $0.15. It went up to $0.30 by Sep 2025 (+100%), then went up to $0.50 by Mar 2026 (+67%), and finally went up to $0.75 by Oct 2026 (+50%). Gemini Flash tier, output has been listed since Mar 2025, starting at $0.60. It went up to $2.50 by Sep 2025 (+317%), then went up to $3 by Mar 2026 (+20%), and finally went up to $3.75 by Oct 2026 (+25%). xAI API: grok-4.7, input appeared by Oct 2026 at $2. grok-4.7, output appeared by Oct 2026 at $6.
Are AI model APIs prices going up or down?
Across the 22 plans we chart, list prices over the last three years are rising, by a median 4% a year. Cuts outnumber increases, 46 to 39.
Where are AI model APIs prices heading?
Our AI price index tracks one line for the whole market, with the statistical trend, outside research and our own scenarios to 2028 and 2030.

Other categories

Paste the address of its pricing page and we will add it to the tracker.