OpenAI's Karan Singhal: model intelligence got 100x cheaper in a year, GPT-6 Luna beats 5.6 Sol at 34x lower cost
FinanceYF5 · x · 2026-10-22
In an a16z interview, OpenAI's Karan Singhal discussed the collapsing cost of intelligence and AI in healthcare.
Key takeaways:
- GPT-6 Luna is OpenAI's cheapest frontier model on the API; it outperforms 5.6 Sol — their best model a month ago — while being 34x cheaper, measured on their health benchmark;
- Most tiers of model intelligence became over 100x cheaper within roughly a year;
- One tier went from $32 to $0.02 per million tokens in two years;
- Over 300M people per week use ChatGPT for health and wellness questions, and Singhal argues this is just the beginning. He also covered how OpenAI builds health intelligence, why general models beat specialized medical ones, and how AI can cut healthcare costs.
Related event: OpenAI says AI intelligence costs have collapsed ~1600x in two years(2 posts)→
More from Models
- Lighthouse Benchmark: Gemini Argon Runs 8x Cheaper Than Claude Opus — cgarciae88 · 2026-10-22
- Users complain Anthropic models have 'suddenly got really stupid' — gaganghotra_ · 2026-10-22
- AXA, Vaisala and others deploy NVIDIA Earth-2 open weather models in production — NVIDIA Blog · 2026-10-22
- The '4' in Gemini 4 was a randomly chosen number, insider says — prajdabre · 2026-10-22
- DeepSeek Search debuts at $0.75/1k searches, cheapest paid Search API, ranks 17th of 27 — ArtificialAnlys · 2026-10-22
- Lafa scales flow language models to 8B, matching AR quality at 2.5x throughput — AdtRaghunathan · 2026-10-22