Newsbite analysis · AI, Startups
The Decoder
OpenAI's GPT-6 Sol and Luna cut prices in half but barely move the needle on performance
OpenAI has launched the GPT-6 Sol and Luna models, cutting their prices by 50% to match their predecessors' performance metrics. Sol costs $2 and Luna $0.10 per million input tokens, positioned against Anthropic's models. Despite these price cuts, benchmarks suggest minimal improvements in intelligence scores compared to GPT-5.6, creating uncertainty about their real-world application. Additionally, performance regressions in some evaluations raise questions about their readiness for complex tasks.
MonitorConfidence: MediumShort term horizonStory: First report
EvidenceOfficial announcementUnverified claimsCompany only
Published 22 Sept 2026, 20:10 UTC6 min read
Lead image — The Decoder
Brief
NewsBite reading:OpenAI launches GPT-6 Sol and Luna with halved prices, performance largely unchanged
TL;DR
- GPT-6 Sol priced at $2 per million input tokens; Luna at $0.10.
- Price cut of 50% compared to GPT-5.6 models.
- Performance metrics show minimal improvements over previous models.
- OpenAI improves prompt caching, offering 90% discount on cached tokens.
- Luna and Sol are designed for cost-effective recurring tasks.
Interesting facts
- OpenAI claims new models are competitive with Anthropic's Claude lineup in terms of price-to-performance.
- Cached input tokens can yield significant cost savings due to improved caching.
- Despite price cuts, some intelligence scores have regressed in key evaluations.