AI Stacks · Price changes checked November 21, 2026

AI Price-Change Tracker and Build-Cost Estimator

Every dated price change and shutdown we've found on AI vendors' own pricing pages, and a calculator that prices your whole build stack before and after each change.

Watch: Episode 22, Let's do the Math! from 03:08. Jordan's LLM Math Roaster calls Gemini, GPT, Claude and Grok through their APIs side by side, the kind of multi-model setup that makes a single vendor's price change easy to absorb.
Subscribe on YouTubeJoin 72K+ subscribers. We build something real with AI every week.

Price-change log

Dated price changes, model removals and API shutdowns, newest first, each from the vendor's own page. We checked every row on November 21, 2026. The first big one is coming up: Gemini 4 Carbon's per-token prices double on December 1, 2026.

EffectiveProductWhat changesSource
Mar 15, 2027
Upcoming
Google: Gemini 2.5 Flash Image (gemini-2.5-flash-image)
Alternatives and prices
Deprecated; scheduled to shut down.
Before: Available in the Gemini API
After: Shut down; Google says migrate to Gemini 3.1 Flash Lite Image
Vendor page
checked November 21, 2026
Dec 1, 2026
Upcoming
Google: Gemini 4 Carbon (gemini-4-carbon), Gemini API paid tier
Alternatives and prices
Input, output and context-caching prices double.
Before: $2.00 in / $10.00 out per 1M tokens; caching $0.10
After: $4.00 in / $20.00 out per 1M tokens; caching $0.20
Vertex AI lists the same change for global endpoints: $2.00 to $4.00 input, $10.00 to $20.00 output (non-global $2.20 to $4.40). (Vertex AI pricing)
Vendor page
checked November 21, 2026
Dec 1, 2026
Upcoming
Google: Gemini 4 Argon (gemini-4-argon), Gemini API paid tier
Alternatives and prices
Input, output and context-caching prices double.
Before: $2.00 in / $10.00 out per 1M tokens; caching $0.10
After: $4.00 in / $20.00 out per 1M tokens; caching $0.20
Vendor page
checked November 21, 2026
Nov 30, 2026
Upcoming
OpenAI: Agent Builder
Alternatives and prices
Product shuts down.
Before: Deprecated, existing users in a transition window
After: Shut down; OpenAI says ChatKit remains available
Vendor page
checked November 21, 2026
Nov 30, 2026
Upcoming
Google: Antigravity free plan (For Individuals)
Alternatives and prices
Claude Sonnet 4.6, Claude Opus 4.6 and GPT-OSS-120b removed from the free plan.
Before: Free plan includes Gemini 3.8 Flash, Claude Sonnet 4.6, Claude Opus 4.6 and GPT-OSS-120b
After: Those three models removed (marked "Will be removed on November 30, 2026")
Vendor page
checked November 21, 2026
Oct 22, 2026
Done
Google: Veo 3.1 Preview in the Gemini API (veo-3.1-generate-preview, -fast-, -lite-)
Alternatives and prices
Preview models shut down in the Gemini API.
Before: Veo 3.1 Preview available in the Gemini API
After: Shut down; Google points to Gemini Omni Flash or the GA Veo models on Vertex AI
Vendor page
checked November 21, 2026
Sep 24, 2026
Done
OpenAI: Sora 2 models and Videos API
Alternatives and prices
API shut down.
Before: sora-2 available through the Videos API
After: Shut down; OpenAI lists no one-to-one replacement API
Vendor page
checked November 21, 2026

What would your stack cost?

Pick the tools you use and enter your monthly usage. The estimator shows your monthly list-price cost at today's prices and again from December 1, 2026, so you can see what the Gemini changes do to your bill. Every change updates the share link.

Rates the estimator uses

Every price below is from the linked vendor page, checked November 21, 2026.

Coding agent

OptionList price usedSource
Cursor Plus$8/mo per seatcursor.com
Cursor Premium$30/mo per seatcursor.com
Cursor Super$100/mo per seatcursor.com
Cursor Ultra$200/mo per seatcursor.com
Claude Pro (Claude Code)$20/mo per seatclaude.com
Claude Max (Claude Code)$100/mo per seatclaude.com
Claude Ultra (Claude Code)$200/mo per seatclaude.com
ChatGPT Plus (Codex)$20/mo per seatlearn.chatgpt.com
ChatGPT Pro (Codex), lowest tier$100/mo per seatlearn.chatgpt.com
GitHub Copilot Pro$10/mo per seatdocs.github.com
GitHub Copilot Pro+$39/mo per seatdocs.github.com
Windsurf Pro$20/mo per seatdevin.ai
Google Antigravity (individual)$0/mo per seatantigravity.google

App builder

OptionList price usedSource
Replit Core$20/moreplit.com
Replit Pro$100/moreplit.com
Lovable Pro$25/molovable.dev
Lovable Business$50/molovable.dev
Bolt Pro$25/mobolt.new
v0 Plus (per user)$30/mov0.app

LLM API

OptionList price usedSource
Gemini 4 Carbon$2 in / $10 out per 1M tokens; $4 / $20 from Dec 1, 2026aistudio.google.com
Gemini 4 Argon$2 in / $10 out per 1M tokens; $4 / $20 from Dec 1, 2026aistudio.google.com
Gemini 3.5 Flash Lite$0.3 in / $2.5 out per 1M tokensaistudio.google.com
Claude Opus 5.5$4 in / $20 out per 1M tokensplatform.claude.com
Claude Sonnet 5.5$2 in / $10 out per 1M tokensplatform.claude.com
Claude Haiku 5.5 (prompts up to 100K)$0.1 in / $0.5 out per 1M tokensplatform.claude.com
gpt-6.2-astra (up to 272K)$8 in / $40 out per 1M tokensdevelopers.openai.com
gpt-6.2-sol$2 in / $10 out per 1M tokensdevelopers.openai.com
gpt-6.1-terra$1 in / $5 out per 1M tokensdevelopers.openai.com
grok-4.7 (under 200K prompt)$2 in / $6 out per 1M tokensdocs.x.ai
grok-4.3 (under 200K prompt)$1.25 in / $2.5 out per 1M tokensdocs.x.ai
deepseek-v4-pro (peak)$1.32 in / $3.96 out per 1M tokensapi-docs.deepseek.com
gpt-oss-120b on Groq$0.15 in / $0.6 out per 1M tokensconsole.groq.com

Voice

OptionList price usedSource
ElevenLabs Starter (about 30 min TTS, then about $0.20/min)$6/moelevenlabs.io
ElevenLabs Creator (about 121 min, then about $0.18/min)$22/moelevenlabs.io
ElevenLabs Pro (about 600 min, then about $0.17/min)$99/moelevenlabs.io
ElevenLabs Agents ($0.08/min, LLM extra)$0.08 per minuteelevenlabs.io
Cartesia voice agents ($0.03/min, LLM extra)$0.03 per minutecartesia.ai
Deepgram Nova-3 streaming speech-to-text ($0.0048/min)$0.0048 per minutedeepgram.com
Vapi hosting ($0.05/min; STT, LLM and voice passed through)$0.05 per minutevapi.ai

Auth and database

OptionList price usedSource
Supabase Free$0/mosupabase.com
Supabase Pro (from)$25/mosupabase.com
Firebase Spark$0/mofirebase.google.com
Clerk Pro (auth only)$25/moclerk.com
Neon Launch (Neon's own typical-spend example)$15/moneon.com

Hosting

OptionList price usedSource
Vercel Hobby$0/movercel.com
Vercel Pro (1 seat)$20/movercel.com
Netlify Personal$9/monetlify.com
Netlify Pro$20/monetlify.com
Cloudflare Workers Paid (minimum)$5/modevelopers.cloudflare.com
Replit Reserved VM (shared 0.5 vCPU)$15/modocs.replit.com
DigitalOcean App Platform smallest container$5/modigitalocean.com

Vector database

OptionList price usedSource
pgvector on your Postgres (no separate charge stated)$0/mosupabase.com
Pinecone Starter$0/mopinecone.io
Pinecone Standard (monthly minimum)$50/mopinecone.io
Weaviate Flex (monthly minimum)$45/moweaviate.io
turbopuffer Launch (monthly minimum)$16/moturbopuffer.com

AI video

OptionList price usedSource
Gemini Omni 1.2 Flash 720p (about $0.07/s)$0.07 per secondaistudio.google.com
Kling 3.0 720p with audio ($0.126/s)$0.126 per secondkling.ai
Seedance 2.0 720p (BytePlus example $0.15/s)$0.15 per seconddocs.byteplus.com
Runway gen4_turbo (5 credits/s at $0.01 = $0.05/s)$0.05 per seconddocs.dev.runwayml.com

Automation

OptionList price usedSource
n8n Community Edition (self-hosted)$0/mon8n.io
Zapier Professional, 750 tasks (monthly billing)$29.99/mozapier.com
Zapier Team, 2,000 tasks (monthly billing)$103.50/mozapier.com
Pipedream Startup (monthly billing)$150/mopipedream.com

What broke when prices and models changed on us

We build with these tools every week on the show. These are the times a price, a usage limit or a retired model caught us out, in our own words.

A model we built on got replaced within weeks

In Episode 17, a week after the Sora 2 API came out, Jordan demoed a standalone tool he'd built on it. It already had rough edges:

When I add a reference image, it has to be this exact size. Otherwise, it doesn't work. And sometimes, like, if you make too long of a of a post of of of an input, then sometimes the the the API fails

— Jordan Metzner, Episode 17, at 03:34

Asked why it wasn't just another option in the team's main video tool, he said:

I probably will and, like, kind of deprecate the other one. I think when Sora two came out, it was so new, and I didn't expect kind of these other models to follow-up so quickly

— Jordan Metzner, Episode 17, at 06:24

OpenAI shut down the Sora 2 API on September 24, 2026 (see the log above). The lesson: put each model behind one switch in your code, so a shutdown or price jump means changing a setting instead of rebuilding a tool.

Usage-based bills that surprised us

we saw recently how one developer used a new feature in cursor and left us with like a $2,000 bill.

— Jordan Metzner, Episode 36, at 13:54

the [Codex] computer use feature over the past few weeks, like, definitely killed us on spend on tokens. And it took us like two and a half weeks to just figure out what it was that was causing all the spend.

— Jordan Metzner, Episode 45, at 21:13

I ran it once like that and like my they like immediately told me like I ran out of tokens.

— Jordan Metzner, Episode 46, at 07:05

None of these were price changes. They're what happens when usage-priced tools meet a new feature. A doubled per-token price, like Gemini 4 Carbon's on December 1, multiplies exactly that kind of spend. Set spending caps where the vendor offers them (Cursor's Teams plans do), and check your top cost driver every week.

How to use this page

  • On Gemini 4 Carbon or Argon? Put your monthly token volumes into the estimator and compare the two columns. Then price the same volumes on another model before December 1.
  • Building on an API that's being shut down? Each log row links to the stack page with current alternatives and their prices.
  • Comparing whole stacks? Load a preset, change one tool at a time, and share the link with your team.

FAQ

When does Gemini 4 Carbon get more expensive?

On December 1, 2026. Google's Gemini API pricing page lists Gemini 4 Carbon at $2.00 per 1M input tokens and $10.00 per 1M output tokens through November 30, 2026, then $4.00 and $20.00 starting December 1, 2026. Context caching goes from $0.10 to $0.20. Gemini 4 Argon has the same change, and Vertex AI lists it for global endpoints too.

How much more will Gemini 4 Carbon cost me?

Twice as much per token. An app that sends 20 million input tokens and 4 million output tokens a month pays $80 at the November price ($40 input plus $40 output) and $160 from December 1 ($80 plus $80). Put your own volumes into the estimator on this page to see your number.

Which models cost what Gemini 4 Carbon used to?

On the vendors' own pages (checked November 21, 2026), Claude Sonnet 5.5 and gpt-6.2-sol are listed at $2 per 1M input and $10 per 1M output tokens, and grok-4.7 at $2 and $6 for prompts under 200K tokens. Same price doesn't mean same quality, so test them on your own workload before you switch.

How accurate is the build-cost estimator?

It multiplies list prices from each vendor's pricing page by the usage you enter. It leaves out taxes, annual-billing discounts, free credits and trials, overage rules it can't model, and anything priced only through a sales call. Use it to compare stacks and to see the effect of a price change, then check the linked pricing pages before you buy.

How do I share my stack?

Every change you make updates the page address with a ?stack= parameter. Click Copy share link and paste it anywhere: whoever opens it sees the same tools and usage numbers.

What goes into the price-change log?

Only changes a vendor has dated on its own pricing, model or deprecation page. Each row gives the date, the before and after, a link to the page and the day we checked it. We don't log rumors or prices from third-party sites.

Sources