Anthropic and OpenAI release lower-cost AI models
Anthropic’s Claude Opus 5.5 and OpenAI’s GPT-6 Sol and Luna cut API costs, with savings claims that depend on workload.
By James Whitfield · Staff Writer
3 min read
Anthropic and OpenAI have released new AI models designed to cost less to run: Claude Opus 5.5, GPT-6 Sol and GPT-6 Luna. The Anthropic OpenAI cheaper models releases give developers lower listed API prices as customers look to contain AI spending and the companies face cheaper open-weight rivals, CNBC reported on Sept. 22.
The announcements emphasize efficiency more than a major jump in capabilities. Ars Technica described the reported improvements as iterative, while both companies presented the models as options for work that needs a balance of speed, capability and operating cost.
How much do Anthropic and OpenAI’s new models cost?
API pricing is generally based on tokens, the units of text a model receives as input and generates as output. A lower per-token rate does not by itself determine the final bill, because workloads can require different numbers of tokens and may include caching or other charges.
- Claude Opus 5.5: $4 per million input tokens and $20 per million output tokens, according to Anthropic’s figures reported by Ars Technica and The Decoder.
- GPT-6 Sol: $2 per million input tokens and $10 per million output tokens, Ars Technica reported.
- GPT-6 Luna: $0.10 per million input tokens and $0.50 per million output tokens, according to Ars Technica.
For a simple example using 1 million input tokens and 1 million output tokens, the listed spend would be $24 for Opus 5.5, $12 for Sol and $0.60 for Luna. That comparison excludes cache reads, cache writes and any other charges, so it is a pricing illustration rather than a prediction of an application’s total cost.
Anthropic’s listed input and output rates for Opus 5.5 are 20% below those of Opus 5, The Decoder reported. Its cache-read rate is $0.20 per million tokens, down 60% from the older model’s $0.50 rate. Anthropic says typical operating costs can be about 40% lower because Opus 5.5 also uses fewer tokens; that broader figure is the company’s estimate, not just its stated price reduction.
OpenAI cut Sol and Luna API prices by 50% compared with GPT-5.6 promotional pricing, CNBC reported. Ars Technica separately reported that the models cost half as much as their predecessors, depending on the comparison. Neither report establishes one universal cost baseline for all OpenAI workloads.
What work are Sol, Luna and Opus 5.5 meant to handle?
CNBC reported that Sol sits below OpenAI’s higher-end Astra model and is intended for more complex work including coding. Luna is aimed at high-volume tasks such as extracting information and summarizing documents. Anthropic positions Opus 5.5 for coding and complex knowledge work, according to Ars Technica.
Anthropic says Opus 5.5 generates output more than 30% faster than Opus 5 and includes safeguards for certain cybersecurity, biology and frontier-model-development requests. The company says flagged requests can be routed to another model; that policy does not independently establish that the new model is safer overall.
The pricing moves arrive as organizations weigh when to use costly frontier systems and when to send work to lower-priced alternatives. CNBC reported that OpenAI and Anthropic face pressure from open-weight providers including Alibaba, Moonshot AI and DeepSeek, while customers seek more cost-effective deployments.
This story draws on original reporting from Ars Technica.