--- title: "AI Token Prices 2026: Cost per Million Tokens and How Far They Fell" description: "AI token prices per million: what a GPT-4-class model cost in 2023 versus today, five controls that cut the bill, and the energy behind cheaper tokens." image: https://www.fiegenbaum.solutions/hubfs/pexels-cluster-images/6385891.jpg --- [Skip to content](https://www.fiegenbaum.solutions/en/blog/dramatic-drop-ai-token-prices-opportunities-challenges-sustainability#main-content) [![Fiegenbaum Solutions Logo](https://www.fiegenbaum.solutions/hubfs/Branding/Fiegenbaum-Solutions_Logo.svg) ![Fiegenbaum Solutions Logo](https://www.fiegenbaum.solutions/hubfs/Branding/Fiegenbaum-Solutions_Logo.svg)](https://www.fiegenbaum.solutions/en/?hsLang=en) - [English](https://www.fiegenbaum.solutions/en/blog/dramatic-drop-ai-token-prices-opportunities-challenges-sustainability) - [Deutsch](https://www.fiegenbaum.solutions/blog/tokenpreise-im-sinkflug-so-passen-sie-ihr-geschaeftsmodell-jetzt-an-handlungsempfehlungen-fuer-unternehmen) - Services - [Climate Risk Assessment](https://www.fiegenbaum.solutions/en/climate-risk-analysis-for-companies) - [VSME Report](https://www.fiegenbaum.solutions/en/vsme-reporting-for-smes) - [ESG Due Diligence VCs & Series A+](https://www.fiegenbaum.solutions/en/esg-due-diligence-vcs-series-a-startups) - [Product Carbon Footprint](https://www.fiegenbaum.solutions/en/product-carbon-footprint-lifecycle-assessment) - [Corporate Carbon Footprint](https://www.fiegenbaum.solutions/en/company-carbon-footprint-methods-tools-best-practice) - Solutions - [Materiality Screening](https://www.fiegenbaum.solutions/en/double-materiality-analysis-definition-process-best-practice#fs-screening) - [Atlas](https://atlas.fiegenbaum.solutions/en) - [Load Profile Check](https://atlas.fiegenbaum.solutions/en/lastgang-check) - [Pricing](https://www.fiegenbaum.solutions/en/esg-climate-consulting-fixed-prices-carbon-footprint-vsme-due-diligence-climate-risk) - [About me](https://www.fiegenbaum.solutions/en/about-me) - [Blog](https://www.fiegenbaum.solutions/en/blog) - [Contact](https://www.fiegenbaum.solutions/en/contact) ###### 8 min read # AI Token Pricing 2025: Cost Trends & Sustainability Impact By: [Johannes Fiegenbaum](https://www.fiegenbaum.solutions/en/blog/author/johannes-fiegenbaum) on 7/29/25, 7:23 PM · Last updated October 11, 2026 [ClimateTech Startups](https://www.fiegenbaum.solutions/en/blog/tag/climatetech-startups) [AI & Digital Sustainability](https://www.fiegenbaum.solutions/en/blog/tag/ai-digital-sustainability) ![Close-up of a red circuit board with electronic components](https://www.fiegenbaum.solutions/hubfs/pexels-cluster-images/6385891.jpg) A million input tokens on a frontier model cost roughly thirty dollars in early 2023. The same class of capability now costs cents, and the cheapest usable tiers sit far below a dollar per million tokens. The collapse is real, but it has not made AI cheap: most bills keep rising, because prompts, context windows and agent loops grew faster than unit prices fell. This page sets out the list prices per million input and output tokens, how far they have fallen since 2023, five controls that actually cut a bill, and what cheaper tokens cost in energy and in sustainability reporting terms. **Table of Contents** 1. [How AI Token Pricing Works: Input and Output Tokens per Million](https://www.fiegenbaum.solutions/en/blog/dramatic-drop-ai-token-prices-opportunities-challenges-sustainability#how-ai-token-pricing-works) 2. [AI Token Price Trajectory 2023 to 2026](https://www.fiegenbaum.solutions/en/blog/dramatic-drop-ai-token-prices-opportunities-challenges-sustainability#ai-token-price-trajectory-2023-to-2026) 3. [Cutting AI Token Costs in Practice](https://www.fiegenbaum.solutions/en/blog/dramatic-drop-ai-token-prices-opportunities-challenges-sustainability#cutting-ai-token-costs-in-practice) 4. [What Cheaper Tokens Cost in Energy, Water and Reporting Terms](https://www.fiegenbaum.solutions/en/blog/dramatic-drop-ai-token-prices-opportunities-challenges-sustainability#what-cheaper-tokens-cost-in-energy-water-and-reporting-terms) 5. [Frequently Asked Questions](https://www.fiegenbaum.solutions/en/blog/dramatic-drop-ai-token-prices-opportunities-challenges-sustainability#frequently-asked-questions) ## How AI Token Pricing Works: Input and Output Tokens per Million Language model APIs bill by the token, not by the request. A token is a fragment of text, in English roughly four characters. Every provider quotes the same unit: price per million tokens, split into an input rate for everything you send and an output rate for everything the model writes back. Output almost always costs more than input, usually three to five times more. The reason is mechanical: input is processed in one pass, while output is generated one token at a time, each step re-reading the whole context. Two further rates matter once volume grows. Cached input, where a context the provider has already seen is re-sent, is billed at a fraction of the normal input rate. Sending a forty-page sustainability report to a model costs roughly 30,000 input tokens. Asking for a two-page structured summary adds perhaps 1,500 output tokens. At an input rate of three dollars and an output rate of fifteen dollars per million, that single analysis costs about eleven cents. ## AI Token Price Trajectory 2023 to 2026 Price trackers publish today's rates. What they rarely show is the slope, which is the part that matters when you are signing a two-year budget. The table below follows published list prices per million tokens from the launch of GPT-4 onwards. | Model (release) | Input per 1M | Output per 1M | | --- | --- | --- | | GPT-4, 8k context (March 2023) | $30.00 | $60.00 | | GPT-4 Turbo (November 2023) | $10.00 | $30.00 | | Claude 3 Opus (March 2024) | $15.00 | $75.00 | | GPT-4o (May 2024) | $5.00 | $15.00 | | GPT-4o mini (July 2024) | $0.15 | $0.60 | | Gemini 1.5 Flash (2024) | $0.075 | $0.30 | | DeepSeek-V3 (December 2024) | $0.27 | $1.10 | Provider list prices in US dollars, standard tier, no batch or cache discount. Prices checked on 4 September 2026. Read down the input column and the fall is about two orders of magnitude in under two years. My position on the slope is that it is the least useful number in the table. In my own document-analysis work the per-token price fell far faster than the cost of finishing a job, because the jobs grew: longer context, retries, reasoning steps that bill as output. Budget against cost per completed task, not against the per-token rate, and re-check list prices quarterly. ## Cutting AI Token Costs in Practice Five controls do most of the work. They are listed in the order I would apply them, cheapest change first. - **Cut prompt length.** Long system prompts and pasted boilerplate are billed on every single call. - **Use context caching.** If the same document or instruction set is sent repeatedly, the cached input rate applies. - **Batch what is not urgent.** Overnight classification and extraction rarely need a synchronous answer, and the batch tier is materially cheaper. - **Route by task, not by habit.** Extraction and classification run well on the smallest tier. Reserve the expensive model for drafting and judgement. - **Cap output.** Set a maximum token limit and ask for structured output. One buyer's note from consulting practice: the cheapest model per million tokens is often not the cheapest per completed task. A weaker model that needs two retries, a longer prompt and a verification pass can cost more than the capable model that answers once. Before switching a workload to a cheaper tier, measure the full cost of a finished output, including the calls that failed. ## What Cheaper Tokens Cost in Energy, Water and Reporting Terms Falling unit prices have not reduced consumption, they have increased it. This is a textbook rebound: when a token becomes a hundred times cheaper, organisations do not spend a hundredth as much, they run a thousand times more tokens. Aggregate demand, energy use and spend all rise while the price per unit collapses, which is why a finance team can watch the rate drop and the invoice grow in the same quarter. The physical side is measurable in orders of magnitude rather than decimals. Agency estimates put a single text query to a large model in the range of a fraction of a watt-hour to a few watt-hours, against roughly a tenth of that for a conventional web search. Data centres could account for a materially larger share of global electricity demand by 2030 than the one to two percent they represent today, and water use for cooling scales with the same curve, drawn where the data centres sit rather than where the queries originate. Multiplied across billions of queries, that is why [AI energy demand](https://www.fiegenbaum.solutions/en/blog/ai-energy-demand-sustainable-data-centers?hsLang=en) has become a grid question for data-centre operators. For a company that buys AI as a service, this is not an abstraction. Purchased AI services sit in the upstream value chain, which places them in Scope 3 rather than in your own energy figure. Organisations preparing a [sustainability report under the VSME standard](https://www.fiegenbaum.solutions/en/blog/sustainability-reporting-vsme-standard-smes?hsLang=en) or tracking [Scope 3 emissions](https://www.fiegenbaum.solutions/en/blog/science-based-targets-initiative-sbti-guide-eu-companies?hsLang=en) should ask providers for energy and location data at contract stage, when there is still leverage, and record AI spend as a separate line in [ESG data management](https://www.fiegenbaum.solutions/en/blog/unlocking-long-term-value-esg-data-business-success?hsLang=en). The direction of travel is clear enough to plan around without forecasting a number: list prices keep falling, total consumption keeps rising. Put the price re-check and the energy question on the same quarterly procurement cycle, and the collapse becomes a budget advantage instead of an unmanaged [climate risk](https://www.fiegenbaum.solutions/en/blog/physical-risks-and-transition-risks-how-to-assess-climate-risks?hsLang=en) exposure. ## Frequently Asked Questions Which model is cheapest per million tokens, and cheapest per completed task? The cheapest per million tokens is always a small or lightweight tier, currently well under a dollar for input. The cheapest per completed task is frequently a mid-tier model, because it answers correctly first time. Measure both before switching a production workload. Why do input and output tokens cost different rates? Input is processed in a single pass over the whole prompt. Output is generated token by token, each new token requiring another pass over the growing context, so providers price it several times higher. Is a subscription plan cheaper than paying per token via API? For individual interactive use, a flat monthly subscription is usually cheaper and simpler. For anything automated or high volume, the API is cheaper and measurable per workload. Most organisations end up with both. Do I pay for tokens in failed or retried requests? If the model produced a response, you pay for it, even when your application discards the result. Requests that fail before generation are generally not billed, but a timeout after partial output usually is. --- *Last reviewed: 4 September 2026. List prices change frequently, check the provider pricing page before budgeting.* Next step Turn climate risk or ESG data into a decision? In a 30-minute intro call we map where your biggest lever is, with no obligation. [Book an intro call](https://www.fiegenbaum.solutions/en/contact?hsLang=en) ![Johannes Fiegenbaum](https://www.fiegenbaum.solutions/hubfs/Johannes_07-modified.png) ### Johannes Fiegenbaum ESG and sustainability consultant based in Hamburg, specialised in VSME reporting, carbon accounting and climate risk analysis. Works with mid-sized companies, investors and financial institutions, and developed the VSME reporting software VSEasy together with clients. [More about](https://www.fiegenbaum.solutions/en/about-me?hsLang=en) Related articles ## You may also like ### [Climate Adaptation & Resilience Hamburg 2040: Pioneering Strict Climate Regulations for Business Transformation Three Strategic Implications for Your Business Update (June 2026), EU context: Hamburg's legally binding 2040 climate target reflects a wider European trend: cities and regions ... Read more →](https://www.fiegenbaum.solutions/en/blog/hamburg-future-referendum-2040-climate-law-business-impact?hsLang=en) ### [ESG Data & Metrics Unlocking Long-Term Value: How ESG Data Drives Business Success Beyond Compliance Read more →](https://www.fiegenbaum.solutions/en/blog/unlocking-long-term-value-esg-data-business-success?hsLang=en) ### [VSME Materiality Assessment for SMEs: From ESG Compliance to Growth Opportunities Read more →](https://www.fiegenbaum.solutions/en/blog/materiality-assessment-smes-esg-compliance-growth?hsLang=en) ### [Carbon Accounting & Scope 3 LCA Methodologies in Practice: A Guide to Sustainable Product Development Read more →](https://www.fiegenbaum.solutions/en/blog/understanding-lca-methodologies-and-standards-a-guide-to-sustainable-product-development?hsLang=en) ### [VSME How to Create a Comprehensive Sustainability Report: Key Steps and Standards Read more →](https://www.fiegenbaum.solutions/en/blog/create-comprehensive-sustainability-report-steps-standards?hsLang=en) ### [VSME Unlocking the Value of Third-Party ESG Verification: A Strategic Guide for Forward-Thinking Companies Read more →](https://www.fiegenbaum.solutions/en/blog/third-party-esg-audits-guide?hsLang=en) ##### Johannes Fiegenbaum Johannes Fiegenbaum I turn climate risk and ESG data into capital and financing decisions. For mid-market companies and investors, data-driven. From Hamburg, for companies across Europe. ##### Tools Tools - [Atlas](https://atlas.fiegenbaum.solutions/en) ##### Services Services - [Climate Risk Assessment](https://www.fiegenbaum.solutions/en/climate-risk-analysis-for-companies) - [VSME Reporting](https://www.fiegenbaum.solutions/en/vsme-reporting-for-smes) - [Company Carbon Footprint](https://www.fiegenbaum.solutions/en/company-carbon-footprint-methods-tools-best-practice) - [Double Materiality Assessment](https://www.fiegenbaum.solutions/en/double-materiality-analysis-definition-process-best-practice) ##### Fiegenbaum Solutions Fiegenbaum Solutions - [About me](https://www.fiegenbaum.solutions/en/about-me) - [Blog](https://www.fiegenbaum.solutions/en/blog) - [Contact](https://www.fiegenbaum.solutions/en/contact) [![1% for the Planet Business Member](https://www.fiegenbaum.solutions/hubfs/1ftp_BusinessMember_Horizontal_FullColor-250x107-faa8874b-96e3-41e2-bbbe-9303832ac91c.png)](https://www.onepercentfortheplanet.org/) - [Legal Notice](https://www.fiegenbaum.solutions/en/legal-notice) - [Privacy Policy](https://www.fiegenbaum.solutions/en/privacy-policy) © 2026 Fiegenbaum Solutions ```json { "@context" : "https://schema.org", "@type" : "BlogPosting", "author" : { "@type" : "Person", "name" : "Johannes Fiegenbaum", "url" : "https://www.fiegenbaum.solutions/en/blog/author/johannes-fiegenbaum" }, "dateModified" : "2026-04-01T15:19:54.179Z", "datePublished" : "2025-07-29T17:23:04.000Z", "headline" : "AI Token Prices 2026: Cost per Million Tokens and How Far They Fell", "image" : [ "https://www.fiegenbaum.solutions/hubfs/pexels-cluster-images/6385891.jpg" ], "mainEntityOfPage" : { "@id" : "https://www.fiegenbaum.solutions/en/blog/dramatic-drop-ai-token-prices-opportunities-challenges-sustainability", "@type" : "WebPage" }, "publisher" : { "@type" : "Organization", "logo" : { "@type" : "ImageObject", "url" : "https://www.fiegenbaum.solutions/hubfs/FS_Logo.svg" }, "name" : "Fiegenbaum & Cie. GmbH" } } ``` ```json { "@context" : "https://schema.org", "@id" : "https://www.fiegenbaum.solutions/#organization", "@type" : "Organization", "areaServed" : { "@type" : "Place", "name" : "European Union" }, "description" : "Nachhaltigkeitsberatung mit Fokus auf Klimarisikoanalyse, ESG-Reporting (CSRD, VSME) und Climate Tech Investmentberatung.", "employee" : { "@id" : "https://www.fiegenbaum.solutions/#founder" }, "founder" : { "@id" : "https://www.fiegenbaum.solutions/#founder" }, "image" : "https://www.fiegenbaum.solutions/hubfs/Branding/Fiegenbaum-Solutions_Logo.svg", "knowsAbout" : [ "ESG Reporting", "CSRD Compliance", "VSME Standard", "Climate Risk Analysis", "EURO-CORDEX", "CMIP6", "Scope 3 Emissions", "Scenario Analysis", "Sustainability Strategy", "Climate Tech Investment" ], "logo" : { "@type" : "ImageObject", "url" : "https://www.fiegenbaum.solutions/hubfs/Branding/Fiegenbaum-Solutions_Logo.svg" }, "name" : "Fiegenbaum Solutions", "sameAs" : [ "https://www.linkedin.com/in/johannesfiegenbaum/" ], "url" : "https://www.fiegenbaum.solutions" } ``` ```json { "@context" : "https://schema.org", "@id" : "https://www.fiegenbaum.solutions/#founder", "@type" : "Person", "affiliation" : [ { "@type" : "Organization", "description" : "Climate tech angel syndicate (Venture Partner)", "name" : "Climate Insiders", "url" : "https://climateinsiders.com/" } ], "alumniOf" : { "@type" : "Organization", "name" : "Universität Tübingen", "url" : "https://uni-tuebingen.de/" }, "description" : "ESG consultant and climate tech investment advisor based in Hamburg. Specializing in VSME reporting, carbon accounting, and climate risk analysis for mid-sized companies, investors, and financial institutions.", "image" : "https://www.fiegenbaum.solutions/hubfs/Johannes_07-modified.png", "jobTitle" : "ESG Consultant & Climate Tech Investment Advisor | CSRD, VSME & Climate Risk Specialist", "knowsAbout" : [ "CSRD", "VSME", "ESG Reporting", "Climate Risk Analysis", "EU Taxonomy", "TNFD", "Climate Tech Investment", "Sustainability Strategy", "Carbon Accounting", "Article 9 Funds" ], "name" : "Johannes Fiegenbaum", "publishingPrinciples" : "https://www.fiegenbaum.solutions/en/legal-notice#editorial", "sameAs" : [ "https://www.linkedin.com/in/johannesfiegenbaum/" ], "url" : "https://www.fiegenbaum.solutions/ueber-mich", "worksFor" : { "@id" : "https://www.fiegenbaum.solutions/#organization" } } ``` ```json { "@context" : "https://schema.org", "@id" : "https://www.fiegenbaum.solutions/#website", "@type" : "WebSite", "inLanguage" : [ "de-DE", "en-EU" ], "name" : "Fiegenbaum Solutions", "publisher" : { "@id" : "https://www.fiegenbaum.solutions/#organization" }, "url" : "https://www.fiegenbaum.solutions" } ``` ```json { "@context" : "https://schema.org", "@type" : "FAQPage", "mainEntity" : [ { "@type" : "Question", "acceptedAnswer" : { "@type" : "Answer", "text" : "The cheapest per million tokens is always a small or lightweight tier, currently well under a dollar for input. The cheapest per completed task is frequently a mid-tier model, because it answers correctly first time. Measure both before switching a production workload." }, "name" : "Which model is cheapest per million tokens, and cheapest per completed task?" }, { "@type" : "Question", "acceptedAnswer" : { "@type" : "Answer", "text" : "Input is processed in a single pass over the whole prompt. Output is generated token by token, each new token requiring another pass over the growing context, so providers price it several times higher." }, "name" : "Why do input and output tokens cost different rates?" }, { "@type" : "Question", "acceptedAnswer" : { "@type" : "Answer", "text" : "For individual interactive use, a flat monthly subscription is usually cheaper and simpler. For anything automated or high volume, the API is cheaper and measurable per workload. Most organisations end up with both." }, "name" : "Is a subscription plan cheaper than paying per token via API?" }, { "@type" : "Question", "acceptedAnswer" : { "@type" : "Answer", "text" : "If the model produced a response, you pay for it, even when your application discards the result. Requests that fail before generation are generally not billed, but a timeout after partial output usually is." }, "name" : "Do I pay for tokens in failed or retried requests?" } ] } ```