DeFi Intel

Mizuho cuts Circle price target to $50 on Open USD margin threat

Mizuho lowered its price target on Circle to $50, citing a margin threat from Open USD.

TL;DR

OpenAI and Anthropic dramatically reduced mid‑tier model prices in mid‑2026, with OpenAI cutting GPT‑5.6 Luna by 80% and Anthropic fixing Claude Sonnet 5 rates. Chinese firms like DeepSeek undercut these prices, intensifying competition and pressuring margins as both U.S. firms eye upcoming IPOs.

Key facts

  • OpenAI cut GPT‑5.6 Luna to $0.20 per million input tokens on July 30, 2026.
  • OpenAI reduced GPT‑5.6 Terra to $2 per million input tokens and $12 per million output tokens.
  • Anthropic set Claude Sonnet 5 permanent pricing at $2 per million input tokens and $10 per million output tokens on August 10, 2026.
  • OpenAI launched GPT‑6 Luna at $0.10 per million input tokens and $0.50 per million output tokens on September 22, 2026.
  • Anthropic introduced Claude Opus 5.5 at $4 per million input tokens and $20 per million output tokens, citing 40% effective savings.
  • Chinese AI firm DeepSeek V4 Flash is priced at roughly $0.14 per million input tokens and $0.28 per million output tokens.

OpenAI and Anthropic have dramatically slashed the prices of their mid‑tier AI models in a coordinated push that threatens the competitiveness of open‑source alternatives. On July 30, OpenAI reduced the cost of its GPT‑5.6 Luna model by 80%, setting rates at $0.20 per million input tokens and $1.20 per million output tokens. A few weeks later, Anthropic made its Claude Sonnet 5 pricing permanent at $2 per million input tokens and $10 per million output tokens. The aggressive pricing wave, amplified by Chinese competitors offering even lower rates, is reshaping developer choices and putting margin pressure on the leading U.S. AI firms.

Background

The AI pricing war entered a new phase in 2026 as the two dominant U.S. providers, OpenAI and Anthropic, began a series of steep price reductions across their model families. Historically, developers have balanced cost against performance, often opting for open‑source models that require self‑hosting but avoid per‑call fees. However, the mid‑tier tier of proprietary models has traditionally commanded a premium for reliability and ease of integration. In the summer of 2026, OpenAI cut the price of its GPT‑5.6 Luna model by 80% and later introduced GPT‑6 Sol and Luna at rates roughly half of the previous generation. Anthropic responded by fixing its Claude Sonnet 5 pricing at $2/$10 per million tokens and later launching Claude Opus 5.5 at $4/$20 list pricing, while touting efficiency gains that effectively lower costs for users. Chinese AI firms, notably DeepSeek, have added pressure by offering models such as V4 Flash at roughly $0.14 per million input tokens and $0.28 per million output tokens, a price point that undercuts the reduced U.S. rates. This competitive landscape forces enterprises to segment workloads: cheap proprietary models handle routine tasks, while premium models are reserved for specialized, high‑performance needs. The pricing battle is also linked to upcoming IPO ambitions for both OpenAI and Anthropic, where higher usage metrics are prized even as per‑call margins shrink.

What happened

On July 30, OpenAI announced an 80% price cut for its GPT‑5.6 Luna model, bringing the cost down to $0.20 per million input tokens and $1.20 per million output tokens, according to CryptoBriefing. The same source reported that OpenAI also reduced the price of its higher‑end GPT‑5.6 Terra model by 20%, setting it at $2 per million input tokens and $12 per million output tokens. Anthropic followed on August 10, making the introductory pricing for Claude Sonnet 5 permanent at $2 per million input tokens and $10 per million output tokens. Later, on September 22, OpenAI launched GPT‑6 Sol and Luna with list prices of $2 per million input tokens and $10 per million output tokens for Sol, and $0.10 per million input tokens and $0.50 per million output tokens for Luna, representing roughly a 50% reduction from the previous generation. Anthropic countered with Claude Opus 5.5, listed at $4 per million input tokens and $20 per million output tokens, while noting built‑in efficiency improvements that deliver around 40% effective savings for users. The source also highlighted that Chinese competitor DeepSeek V4 Flash is priced at roughly $0.14 per million input tokens and $0.28 per million output tokens, undercutting the reduced U.S. pricing. The article noted that both OpenAI and Anthropic are expected to pursue IPOs, and that the price cuts are intended to boost adoption and usage metrics even as they compress per‑call margins.

Market impact

The price cuts have reshaped the cost landscape for AI developers. OpenAI's GPT‑5.6 Luna now costs $0.20 per million input tokens and $1.20 per million output tokens, a stark contrast to its previous pricing. GPT‑5.6 Terra's new rates of $2 input and $12 output per million tokens make it more affordable for higher‑end workloads. Anthropic's Claude Sonnet 5 permanent pricing of $2/$10 per million tokens provides a stable cost baseline for users. The September launch of GPT‑6 Luna at $0.10/$0.50 per million tokens represents the lowest price point among the major U.S. providers. Meanwhile, DeepSeek V4 Flash's $0.14/$0.28 per million token pricing from China remains the cheapest option, pressuring U.S. firms to justify higher rates through performance or ecosystem advantages. The combined effect is a shift toward using cheap proprietary models for routine tasks, while open‑source solutions may lose volume unless they can match the new price points. The aggressive pricing also signals that both OpenAI and Anthropic are willing to accept thinner margins in anticipation of higher usage volumes that could support upcoming IPO valuations.

Why it matters

The aggressive pricing strategy matters because it directly influences developer adoption decisions and the competitive dynamics between proprietary and open‑source AI models. Lower costs for mid‑tier models make it easier for enterprises to allocate budget to routine AI tasks without the overhead of self‑hosting, potentially reducing the market share of open‑source alternatives that rely on community contributions and self‑managed infrastructure. The presence of Chinese competitors offering even lower rates intensifies the pressure on U.S. firms to justify higher prices through performance, reliability, or ecosystem benefits. Additionally, the price cuts are tied to the IPO ambitions of OpenAI and Anthropic; higher usage metrics can improve revenue forecasts and investor confidence, even as per‑call margins shrink. This dynamic could reshape funding flows, talent allocation, and the broader trajectory of AI development across borders.

DeFi Intel’s take

Editorial analysis, not a statement of fact.

DeFi Intel's take: the price war sparked by OpenAI and Anthropic suggests a strategic pivot toward volume‑driven growth ahead of their anticipated IPOs. By slashing rates to single‑digit dollar amounts per million tokens, the firms appear to be betting that lower barriers to entry will expand their user bases faster than the margin erosion will hurt profitability. We read this as a signal that the U.S. AI market is preparing for a more commoditized future, where cost efficiency becomes a primary differentiator alongside raw performance. The Chinese pricing undercut underscores a global competitive pressure that could force the leading providers to further innovate or consolidate. If the trend continues, developers may increasingly favor cheap, high‑volume proprietary APIs for routine workloads, while open‑source projects will need to either lower costs or demonstrate clear superiority in niche applications to retain relevance.

Sources used for this analysis

DeFi Intel wrote this analysis from the reporting above. Any figure or quotation can be checked against those pages.

Frequently asked questions

What specific price reductions did OpenAI implement in July 2026?

OpenAI cut the price of its GPT‑5.6 Luna model by 80%, setting the cost at $0.20 per million input tokens and $1.20 per million output tokens. It also reduced the price of its GPT‑5.6 Terra model by 20%, bringing it to $2 per million input tokens and $12 per million output tokens.

How does Anthropic's pricing compare after its September update?

Anthropic introduced Claude Opus 5.5 with a list price of $4 per million input tokens and $20 per million output tokens, while noting that built‑in efficiency improvements deliver around 40% effective savings for users compared with earlier rates.

Why are Chinese AI models influencing the U.S. pricing war?

Chinese competitor DeepSeek V4 Flash offers pricing of roughly $0.14 per million input tokens and $0.28 per million output tokens, undercutting even the reduced U.S. rates and forcing American providers to reconsider how they price their models to stay competitive.

Direct answers

Short answers, written to be quoted.

What are the new prices for OpenAI's GPT‑5.6 Luna model?
The new price for OpenAI's GPT‑5.6 Luna model is $0.20 per million input tokens and $1.20 per million output tokens after an 80% cut.
How does Anthropic's Claude Sonnet 5 pricing compare to its previous rates?
Anthropic made Claude Sonnet 5 pricing permanent at $2 per million input tokens and $10 per million output tokens, establishing a stable cost structure.
Why are Chinese AI models affecting U.S. pricing strategies?
Chinese AI models like DeepSeek V4 Flash are priced at roughly $0.14 per million input tokens and $0.28 per million output tokens, undercutting U.S. rates and prompting American providers to lower prices.
Read full article on Gnews_circle_usdc →

DeFi Intel is an entity-graph aggregator: we curate, tag and link crypto news to a typed knowledge graph of protocols, tokens, people and incidents. We do not republish the full article body. Use the link above to read the original report at Gnews_circle_usdc.

Entities in this story

Discussion

Comments are reviewed before publication.

No comments yet.

Stay current on entity-tagged crypto news

Get the weekly DeFi Intel brief — entity-graph intelligence covering protocols, tokens and incidents, free to your inbox.