Anthropic and OpenAI Cut Prices With Claude Opus 5.5, GPT-6 Sol and GPT-6 Luna

Anthropic and OpenAI each announced cheaper AI models on Tuesday, responding to mounting pressure from lower-cost open-weight competitors out of China and elsewhere.

Anthropic unveiled Claude Opus 5.5, the newest entry in its Claude 5.5 family. According to Anthropic, the model is more token-efficient and will cost approximately 40% less to run than Claude Opus 5. Dianne Penn, Anthropic's head of product management, research and labs, told CNBC the company has focused on making the model's reasoning process use fewer tokens depending on a user-configurable "effort setting." Anthropic did not disclose exact per-token pricing.

OpenAI introduced two new tiers to its GPT-6 lineup: GPT-6 Sol and GPT-6 Luna. Sol sits one tier below OpenAI's existing Astra model and is positioned for complex workloads such as coding. Luna is aimed at high-volume tasks like document summarization and information extraction. OpenAI says API pricing for both models is 50% lower than the promotional pricing offered for GPT-5.6. The company has not published absolute per-1M-token rates for either model.

Neither company disclosed context window sizes, parameter counts, or benchmark scores for the new releases.

First releases since the slowdown call

These launches are the first from either lab since Anthropic CEO Dario Amodei called for an industrywide slowdown on advanced AI development amid intensifying safety debate. The controversy escalated after former Anthropic researcher Jacob Coxon posted on X on September 8 that he had quit his job, warning that both labs were "gambling with our lives." OpenAI CEO Sam Altman and Tesla/SpaceX CEO Elon Musk subsequently voiced support for a slowdown as well.

Despite that rhetoric, both companies moved forward with new commercial releases within weeks, underscoring the gap between public safety statements and competitive product pressure.

Competitive pressure from open-weight models

The price cuts arrive as OpenAI and Anthropic contend with enterprise customers actively trying to reduce AI spending, and as open-weight models from Alibaba, Moonshot AI, and DeepSeek continue to undercut frontier lab pricing while closing performance gaps on many benchmarks. Both Sol/Luna and Opus 5.5 appear designed to give cost-sensitive customers a mid-tier option without abandoning the higher-priced flagship models (GPT-6 Astra and Claude Opus 5) that remain positioned at the top of each company's lineup.

What this means

The timing is notable: two labs whose leadership publicly called for slowing AI development shipped new commercial products just weeks later. That's not necessarily hypocrisy — the new models are explicitly cheaper, smaller-tier variants rather than more capable frontier systems — but it does illustrate that competitive dynamics with open-weight rivals are driving release cadence more than the safety debate is slowing it. Enterprises watching AI costs should note that neither company has published hard pricing numbers yet, only relative discounts against prior promotional rates, making real cost comparisons difficult until official rate cards appear.