OpenAI launches cheaper GPT-6 Sol and Luna models

OpenAI on Tuesday announced GPT-6 Sol and GPT-6 Luna, two new models it describes as cheaper and faster additions to the GPT-6 Astra family, according to The Information. The release marks an effort by the company to broaden access to its latest generation of AI beyond the flagship Astra model, which launched earlier this month at premium pricing.techcrunch

Filling the Pricing Gap

The new models follow the tiered structure OpenAI established with its GPT-5.6 family in July, where Sol served as the flagship, Terra as a mid-tier option, and Luna as the fastest and most affordable. GPT-6 Sol and Luna replace their GPT-5.6 predecessors, with OpenAI claiming improvements in professional work, coding, computer use, and accuracy.openai+1

GPT-6 Astra, which rolled out starting September 3, is priced at $10 per million input tokens and $50 per million output tokens on the API — a rate that matched Anthropic's competing Fable 5.1 model but represented a steep jump from GPT-5.6 Sol's promotional rate of $4 input and $20 output. The New Stack reported that OpenAI "halved GPT-6's price" with the new models in a move to undercut Anthropic on cost. Reuters Thomson Reuters Corporation ↗ 2.92% noted the launch of the "low-cost" Sol and Luna models follows the Astra rollout earlier this month.cnbc+3

Competitive Pressure and Timing

The announcement comes on the same day Anthropic unveiled Claude Opus 5.5 and amid broader price competition in the AI industry. The Information separately reported Tuesday that large software firms are offering AI discounts to counter both Anthropic and OpenAI. Microsoft ↗ 0.52%, which distributes OpenAI's models through its Azure cloud platform, and Amazon Amazon.com, Inc. ↘ 2.24%, which offers them via Bedrock, stand to benefit from a wider range of price points for enterprise customers.marketscreener

OpenAI's DevDay 2026 is scheduled for September 29 in San Francisco, just one week after Tuesday's release. The early launch may be a strategic move to clear the way for additional announcements at the developer conference. ZDNet noted the new models arrive "less than three months after GPT-5.6," underscoring the accelerating pace of model releases across the industry.zdnet+1