OpenAI Launches GPT-6 Sol and Luna, API Pricing Cut 50% from Previous Generation
nashnova research
OpenAI released GPT-6 Sol and Luna on September 22, halving API prices versus the prior generation — just 90 minutes after Anthropic shipped Opus 5.5. AI models have entered a cycle of rising performance and falling prices, reshaping downstream usage and spending.
What do the two new models actually do?
Sol targets complex tasks like coding. Luna handles high-frequency, goal-clear work — document summaries, information extraction, quick Q&A.
This means → OpenAI is splitting the GPT-6 line into "heavy" and "light" tracks, so users stop paying top-tier compute for simple jobs.
Sol's factual-accuracy error rate is roughly half that of the prior generation, matching flagship Astra-level reliability — but at far lower cost.
How did they cut prices in half?
API pricing is 50% below the previous 5.6-series Sol and Luna. OpenAI credits caching and inference-efficiency improvements.
In plain terms = they're not selling at a loss — the same GPU now produces more answers per dollar, so each response costs less to serve.
Luna is also available on desktop to free and Go users, rolling out to all users within the day — the paywall drops further.
Why does the release timing matter?
Anthropic launched Opus 5.5 roughly 90 minutes before OpenAI's announcement — near-simultaneous releases.
OpenAI's blog post directly cited benchmark data, claiming Sol and Luna significantly outperform Anthropic's Fable and Opus lines across multiple tasks.
This reflects a release race now measured in hours, not weeks. Product windows have compressed to almost nothing.
What else happened in AI this week?
In the same week, xAI shipped Grok 4.7 and several Chinese labs released new models — all priced well below levels seen just months ago.
This means → the per-token cost decline is not an OpenAI story; it is an industry-wide trend accelerating across every major lab.
A Citadel Securities note to clients stated: falling costs are driving higher usage volumes, overall AI spending is still rising, and the trend points to higher long-term profit potential.
What does the price war mean for the market?
The critical condition: labs must sustain revenue growth through a sustained price war — whether volume can compensate for price determines the profit trajectory.
In plain terms = the price war benefits users, but for labs it is a high-stakes bet — win if usage scales fast enough, lose if it doesn't.
This signals that the AI industry sits at a pivotal "volume-for-price" inflection — spending rises in the short term, but long-term divergence among players is inevitable.
市场有风险，内容仅供研究参考，不构成投资建议。