AI Token Prices Hit Record Lows as Costs Plunge 43%

A closely followed measure of AI token prices touched fresh lows this week, the latest sign of deflating costs across an increasingly competitive AI landscape. The LLM Token Expenditure Index, a daily price gauge from intelligence firm Silicon Data, fell to 97 cents on Monday, its lowest reading since the index was created late last year and more than half its high recorded earlier this summer, according to CNBC's reporting on the decline.

The broader pricing trend behind this single-day low is genuinely dramatic. The price of processing AI tokens fell to roughly $1.16 to $1.18 per million tokens in early August 2026, a 43% decline from $2.04 just at the end of May, according to Cryptobriefing's analysis of the trend. Put in longer-term context, frontier AI intelligence is now priced at approximately 12% of its March 2023 levels, with some benchmarks showing comparable capability costing more than 280 times less than it did in early 2023.

AI Token Pricing Timeline

Two distinct competitive pressures are compounding to push prices lower, and neither shows signs of easing. The first is OpenAI's own decision to slash pricing on its GPT-5.6 Luna models by 80% in late July 2026, bringing input tokens down to $0.20 per million and output tokens to $1.20 per million, according to Cryptobriefing's reporting. The second is the wave of aggressively priced open-weight models emerging from China, with providers like DeepSeek achieving price reductions of up to 99% earlier in 2026, a trend we've tracked closely, including Kimi K3's own open-weight release undercutting Western labs on price.

This price collapse comes with a genuine counterintuitive twist worth understanding directly. ARK Invest describes the current dynamic as a "virtuous cycle," where lower prices drive broader adoption, which drives higher usage volumes, which justifies further infrastructure investment, which in turn enables even lower prices, according to Cryptobriefing's reporting. Token processing volumes have reportedly surged up to tenfold in certain agent-driven applications specifically, since AI agents automatically generate far more token consumption per task than a single human-typed query ever would.

That volume surge means falling per-token prices don't automatically translate into falling total AI spending for businesses. Google offers the cleanest public evidence of this pattern: at I/O 2026, CEO Sundar Pichai said Google's token processing grew from 9.7 trillion tokens a month to more than 3.2 quadrillion in two years, a roughly 330-fold increase, even as the cost of GPT-4-level quality output fell from $30 per million tokens to under $0.50, according to an industry analysis of the trend. Palo Alto Networks CEO Nikesh Arora has separately argued token costs still need to fall as much as 90% further to unlock genuinely large-scale enterprise AI adoption.

This pricing trend is worth understanding directly for any business budgeting for AI infrastructure or evaluating vendor contracts, since the headline story, "AI is getting cheaper," only tells half the story. As this data shows, businesses scaling AI agent deployments specifically should expect total spending to potentially rise even as per-token costs fall, since agentic workflows tend to consume dramatically more tokens than simple chat-based usage.

For businesses negotiating AI vendor contracts, this competitive price war between OpenAI's aggressive cuts and Chinese open-weight alternatives is a genuine opportunity to renegotiate pricing or test cheaper alternatives for high-volume, lower-complexity tasks, while still budgeting for total spend growth as usage scales alongside falling unit costs.

Why are AI token prices falling so quickly?
Two forces are driving the decline simultaneously: OpenAI cut pricing on its GPT-5.6 Luna models by 80% in late July 2026, and competing Chinese open-weight AI models, including from DeepSeek, have achieved price reductions of up to 99%, creating an intense price war among AI providers.

Does falling token pricing mean businesses will spend less on AI?
Not necessarily. Token processing volumes have surged, in some agent-driven applications by up to tenfold, meaning total AI spending can rise even as per-token prices fall, since businesses use dramatically more tokens once costs decrease enough to justify broader automation.

How much have AI token prices fallen since 2023?
Frontier AI intelligence is now priced at roughly 12% of its March 2023 levels, with some benchmarks showing comparable AI capability costing more than 280 times less than it did just three years earlier.

AI token prices hit fresh record lows, with the LLM Token Expenditure Index falling to 97 cents, more than half its summer high, driven by OpenAI's 80% price cut on its GPT-5.6 Luna models and aggressive pricing from Chinese open-weight competitors like DeepSeek. Despite falling per-token costs, total AI spending for many businesses continues rising, since token processing volumes have surged, in some agent-driven applications by up to tenfold, as cheaper pricing unlocks broader, higher-volume AI usage. Executives including Palo Alto Networks CEO Nikesh Arora argue prices still need to fall significantly further, potentially another 90%, before AI adoption reaches its full potential scale.