Calculator
Tools

AI API price changes

87 logged changes, newest first: 18 price cuts, 6 price rises, 36 launches, 11 retirements, 5 tier changes, 11 data corrections. Cuts show in blue with ▼, rises in red with ▲.

Prices checked 10 Oct 2026

Every logged change

Type of change

Showing 87 of 87 changes

October 2026

  • Anthropic launches Claude Haiku 5.5 at $0.10/$0.50 for prompts up to 100K tokens, a tenth of Haiku 4.5's $1/$5. A prompt over 100K tokens, cache reads and writes included, bills the whole request at $0.50/$2.50. 1M context, 128K max output, 50% batch discount. It ties GPT-6 Luna on short prompts; Haiku 4.5 moves to legacyClaude pricingLaunch
  • Anthropic halves Claude Sonnet 5.5 cache reads from $0.20 to $0.10/MTok (0.05x input, the same multiplier as Opus 5.5). Token rates and cache writes are unchanged; Sonnet 5.5 now matches GPT-6.1 Sol's $0.10 cached inputClaude pricing▼ Price cut
  • Azure GPT-5.5 and Azure GPT-5.4 have a 1.05M-token context window, not 1M as we listed.Azure OpenAI pricingCorrection
  • The Manus free plan refreshes 300 credits a day, capped at 1,500 a month, with no starter pack.Tier change
  • DeepSeek's off-peak rate now also covers Chinese public holidays in full.DeepSeek pricingTier change
  • OpenAI deprecates GPT-5.4 Nano, with its API shutting down on Apr 1, 2027 and GPT-6 Luna as the recommended replacementOpenAI pricingRetired
  • DeepInfra retires Llama 4 Maverick, so the tracked rate moves to Novita at $0.27/$0.85 (from $0.20/$0.80), keeping the full 1M contextLlama pricing▲ Price rise

September 2026

  • OpenAI launches GPT-6.1 Sol at $2/$10 (1.05M context; cached input $0.10), replacing GPT-6 Sol a week after launch and matching Claude Sonnet 5.5's price exactlyOpenAI pricingLaunch
  • Anthropic launches Claude Sonnet 5.5 at $2/$10 (1M context), the same price as the Sonnet 5 it replacesClaude pricingLaunch
  • Perplexity ends Sonar Chat Completions support. Sonar moves to the Agent API at $0.25/$2.50 (from $1/$1) with web search billed separately at $2.50 per 1,000 calls; Sonar Pro, Reasoning Pro and Deep Research become Agent API presets billed at the underlying model's rates, and are no longer trackedPerplexity Sonar pricingTier change
  • Anthropic launches Claude Opus 5.5 at $4/$20 (1M context), 20% below Opus 5 on both input and output, with cache reads at $0.20/MTok (0.05x input) and fast mode at $8/$40. Opus 5 stays available at $5/$25Claude pricingLaunch
  • OpenAI launches GPT-6 Luna at $0.10/$0.50 (1.05M context), its cheapest current-generation model, alongside GPT-6 Sol at $2/$10OpenAI pricingLaunch
  • xAI launches Grok 4.7 as its new flagship at $2/$6 (500K context; $4/$12 above 200K prompt tokens), the same rates as Grok 4.6 including $0.50 cached input. No batch discountGrok pricingLaunch
  • Google limits the Gemini 2.5 models to developers who have used them before. They are not deprecated and keep their prices, but new projects are steered to Gemini 3.5 Flash-Lite and 3.8 FlashGemini pricingTier change
  • Perplexity will support its Sonar Chat Completions models only until Sep 27, 2026, as it moves users to its Agent APIPerplexity Sonar pricingRetired
  • Codestral's context window is 128K, not 256K (our listing was wrong).Mistral pricingCorrection
  • GPT-5.5, GPT-5.5 Pro and GPT-5.4 have a 1.05M-token context window, not 1M as we listed.OpenAI pricingCorrection
  • Azure retires o4-mini on 19 Nov 2026, not 23 Oct as we listed.Azure OpenAI pricingCorrection
  • DeepSeek V4.1 Flash replaces V4 Flash and cuts Flash rates to $0.30/$1.20 at peak ($0.15/$0.60 off-peak). V4 Pro's planned Sep 14 retirement is reversedDeepSeek pricing▼ Price cut
  • OpenAI launches GPT-6 Astra, its most capable model, at $10/$50 (1.05M context; $20/$75 above 272K; 50% batch discount). Azure Foundry lists it at the same Global Standard rate, without batch pricingOpenAI pricingAzure OpenAI pricingLaunch
  • Google launches Gemini 3.8 Flash GA at an introductory $0.75/$3.75 (1M context), with the $1.50/$7.50 list price from Jan 1, 2027Gemini pricingLaunch
  • Meta releases Muse Spark 1.3 at an unchanged $1.25/$4.25; the tracked Muse Spark row moves from 1.1 to 1.3. Muse Spark 1.2 shipped Aug 5 at the same price and widened access beyond the USLlama pricingLaunch
  • Anthropic launches Claude Fable 5.1 at the same $10/$50 as Fable 5, with cache reads cut 75% to $0.25/MTok; Fable 5 moves to legacyClaude pricingLaunch

August 2026

  • OpenAI puts GPT-5.6 Sol on promotional pricing of $4/$20 (from $5/$30; cached input $0.40), available at least through Nov 21, 2026OpenAI pricing▼ Price cut
  • DeepSeek raises prices and starts peak/off-peak billing, with off-peak at half the peak rate (peak is 01:00-04:00 and 06:00-10:00 UTC on weekdays). Peak rates: V4 Flash $0.44/$1.32 (from $0.14/$0.28) and V4 Pro $1.32/$3.96 (from $0.435/$0.87). DeepSeek is now tracked at peak ratesDeepSeek pricing▲ Price rise
  • Google halves Gemini 3.6 Flash to an introductory $0.75/$3.75 through Dec 31, 2026 (the $1.50/$7.50 list price returns Jan 1, 2027) and ships Gemini 3.7 Flash GA at the same introductory rateGemini pricing▼ Price cut
  • xAI launches Grok 4.6 as its new flagship at $2/$6 (500K context; $4/$12 above 200K prompt tokens), the same token rates as Grok 4.5 but with cached input at $0.50 instead of $0.30. No batch discountGrok pricingLaunch
  • Anthropic makes Claude Sonnet 5's $2/$10 launch rate permanent — the scheduled Sep 1 increase to $3/$15 will not occur. Sonnet 5 was previously tracked at its $3/$15 list priceClaude pricing▼ Price cut
  • Meta hosted rates re-baselined against DeepInfra: Llama 3.3 70B falls to $0.10/$0.32 (-57% input) and Llama 3.1 8B to $0.02/$0.04, the cheapest rate trackedLlama pricing▼ Price cut
  • Meta hosted rates rise for the Llama 4 pair: Maverick to $0.20/$0.80 (+33% input) and Scout to $0.10/$0.30 (+25% input)Llama pricing▲ Price rise
  • DeepSeek announces 2x peak-hour pricing for 09:00-12:00 and 14:00-18:00 Beijing time; start date not yet confirmed, so listed rates are unchangedDeepSeek pricingTier change

July 2026

  • Mistral schedules Magistral Medium 1.2 for retirement; reasoning folds into Mistral Medium 3.5 / Small 4 via reasoning_effortMistral pricingRetired
  • Mistral retires Magistral Medium 1.2, Magistral Small 1.2 and Devstral 2 on the same day. Magistral Medium left the comparison table then; Magistral Small (added in error on Aug 4) and Devstral 2 were removed from our tables in September 2026Mistral pricingRetired
  • OpenAI cuts GPT-5.6 Luna 80% ($1/$6 to $0.20/$1.20) and GPT-5.6 Terra 20% ($2.50/$15 to $2/$12), three weeks after the family's GA. Luna is now OpenAI's cheapest model and Terra undercuts GPT-5.4 on both input and outputOpenAI pricing▼ Price cut
  • Perplexity Sonar Deep Research now tracked at $2/$8, plus $2/MTok citation tokens, $3/MTok reasoning tokens and $5 per 1,000 searchesPerplexity Sonar pricingLaunch
  • DeepSeek retires R1-0528 from its first-party API (deepseek-reasoner alias deprecates 2026-07-24); removed from the comparison tableDeepSeek pricingRetired
  • Anthropic launches Claude Opus 5 ($5/$25, 1M context) — new Opus flagship at the same price as Opus 4.8, with fast mode at $10/$50 and a 512-token prompt-cache minimumClaude pricingLaunch
  • OpenAI confirms o4-mini API shutdown for 2026-10-23; recommended migration is GPT-5.6 Terra. Still listed at $1.10/$4.40 on the standard pricing table until thenOpenAI pricingAzure OpenAI pricingRetired
  • Google launches Gemini 3.6 Flash ($1.50/$7.50, 1M context) — same input as 3.5 Flash with output cut from $9 to $7.50 — and Gemini 3.5 Flash-Lite ($0.30/$2.50) GAGemini pricingLaunch
  • Mistral adds Devstral 2, an open-weight agentic coding model at $0.40/$2.00 (256K context), alongside the Mistral Vibe CLIMistral pricingLaunch
  • OpenAI GPT-5.6 family GA - Sol $5/$30, Terra $2.50/$15, Luna $1/$6 (1.05M context). New cache-write charge at 1.25x input; cached reads 10% of input.OpenAI pricingLaunch
  • Meta opens its first paid first-party API — Muse Spark 1.1 at $1.25/$4.25 (1M context), US developers only in public previewLlama pricingLaunch
  • xAI launches Grok 4.5 - new flagship at $2/$6 (500K context; $4/$12 above 200K prompt tokens). Cached input $0.30, no batch discount, not yet available in the EU.Grok pricingLaunch
  • Anthropic Claude Sonnet 5 now tracked at its effective $2/$10 rate, launched as introductory pricing and made permanent on Aug 11, 2026; adaptive thinking on by defaultClaude pricingLaunch
  • OpenAI adds GPT-5.5 Pro ($30/$180, 1M context) — highest-precision flagship variantOpenAI pricingLaunch
  • Azure adds GPT-5.5 ($5/$30, 1M context) to Foundry Global StandardAzure OpenAI pricingLaunch
  • Llama 3.3 70B: commodity host rates fall to ~$0.23/$0.40 (DeepInfra); $0.88 is now Together-onlyLlama pricing▼ Price cut
  • Meta budget slot updated to Llama 3.1 8B — 128K context, ~$0.03/$0.06 on hosted providers (supersedes Llama 3 8B / 8K)Llama pricingLaunch
  • Mistral Small 4 is $0.15/$0.60 per MTok with a 256K-token context window, not $0.10/$0.30 and 128K as we listed (our listing was wrong).Mistral pricingCorrection
  • Gemini 3.1 Pro and Gemini 2.5 Pro have a 1M-token context window (our listing was wrong).Gemini pricingCorrection
  • GPT-5.4 mini, GPT-5.4 nano, Azure GPT-5.2 and Azure GPT-5-mini have a 400K-token context window (our listing was wrong).OpenAI pricingAzure OpenAI pricingCorrection
  • Azure GPT-5.4 has a 1M-token context window (our listing was wrong).Azure OpenAI pricingCorrection
  • Magistral Medium 1.2 has a 128K-token context window (our listing was wrong).Mistral pricingCorrection
  • OpenAI o4-mini marked legacy — superseded by GPT-5 mini; still served on the APIOpenAI pricingRetired

June 2026

  • Anthropic retires legacy Claude Opus 4 and Sonnet 4 — recommended replacements are Opus 4.8 and Sonnet 4.6Retired
  • Anthropic launches Claude Fable 5 ($10/$50, 1M context) — public Mythos-class model, new top tier above OpusClaude pricingLaunch
  • o4-mini is $1.10/$4.40 per MTok, not $4/$16 as we listed (our listing was wrong).OpenAI pricingCorrection
  • Gemini 2.0 Flash shutdown — removed from the comparison tableGemini pricingRetired

May 2026

  • Mistral: Codestral $0.4/$2→$0.3/$0.9, Mistral Small 4 $0.15/$0.6→$0.1/$0.3; Large 3 context to 256KMistral pricing▼ Price cut
  • Mistral adds Medium 3.5 ($1.50/$7.50) and Magistral Medium 1.2 reasoning model ($2/$5)Mistral pricingLaunch
  • Llama 4 Maverick hosted rates fall to ~$0.15/$0.60 on DeepInfra/OpenRouterLlama pricing▼ Price cut
  • Azure adds GPT-5.4 ($2.50/$15, 272K) as its newest hosted GPT; GPT-5.2 moves to previous-genAzure OpenAI pricingLaunch
  • Anthropic releases Claude Opus 4.8 — same $5/$25 price; new tokenizer carried over from 4.7Claude pricingLaunch
  • DeepSeek makes the V4 Pro 75% price cut permanent — $0.435/$0.87 is now the standard rateDeepSeek pricing▼ Price cut
  • Google launches Gemini 3.5 Flash ($1.50/$9) at I/O 2026; Gemini 3.1 Flash-Lite ($0.25/$1.50) GAGemini pricingLaunch
  • xAI retires Grok 4.1 Fast, Grok 4 and Grok 3 — slugs now redirect to Grok 4.3Grok pricingRetired
  • Grok 4.3 gets API token pricing ($1.25/$2.50, 1M context); xAI ships the Grok Build 0.1 coding modelGrok pricingLaunch
  • Grok 4.20: input $2→$1.25, output $6→$2.50 (matches Grok 4.3); context revised 2M→1MGrok pricing▼ Price cut

April 2026

  • Gemini 2.0 Flash deprecated — full shutdown June 1, 2026Gemini pricingRetired
  • Gemini 2.5 Flash-Lite GA at $0.10/$0.40Gemini pricingLaunch
  • Gemini 3.1 Pro Preview and Gemini 3 Flash Preview now availableGemini pricingLaunch
  • DeepSeek cache-hit input drops to $0.0028/MTok (one-tenth of launch rates)DeepSeek pricing▼ Price cut
  • DeepSeek releases V4 Flash and V4 Pro with 1M context windowDeepSeek pricingLaunch
  • DeepSeek V4 Flash launches at $0.14/$0.28 — 50% cheaper than V3/R1's unified pricingDeepSeek pricing▼ Price cut
  • OpenAI launches GPT-5.5 ($5/$30, 1M context) — new flagship above GPT-5.4OpenAI pricingLaunch
  • xAI launches Grok 4.3 Beta on SuperGrok Heavy ($300/mo subscription, no API token pricing yet)Launch
  • Anthropic releases Claude Opus 4.7 — same $5/$25 price but new tokenizer uses up to 35% more tokensClaude pricingLaunch
  • Meta releases Llama 4 Scout (10M context) and finalises MaverickLlama pricingLaunch
  • Google releases Gemma 4 open-source family under Apache 2.0 (E2B/E4B/26B/31B)Launch
  • Gemini 2.5 Flash: output now includes thinking tokens ($0.60 → $2.50/MTok)Gemini pricing▲ Price rise
  • Gemini 2.0 Flash is $0.10/$0.40 per MTok, not $0.075/$0.15 as we listed (our listing was wrong).Gemini pricingCorrection
  • DeepSeek: unified pricing — V3 & R1 both now $0.28/$0.42, R1 output -81%DeepSeek pricing▼ Price cut
  • Llama 3.3 70B: reverted to $0.88/$0.88 on Together.aiLlama pricing▲ Price rise

March 2026

  • Haiku 4.5: input +25%, output +25%Claude pricing▲ Price rise
  • Gemini 2.0 Flash: input -25%, output -25%Gemini pricing▼ Price cut
  • Mistral Large: input -17%, output -17%Mistral pricing▼ Price cut
  • Llama 3.3 70B: input -19%, output -19%Llama pricing▼ Price cut

Next steps

Price history explorerChart how any model's price has moved.AI API Pricing October 2026: What Changed This MonthThis month's roundup of price moves.API pricing by providerCurrent prices for every provider's models.