[AINews] GPT 5.6 price cut by 20%-80%: Cost of GPT 5.4 Intelligence dropped 13x in 4 months due to GPT 5.6 recursive self-optimization
Latent Space says OpenAI is now selling GPT-5.4-level performance through Luna at roughly one-thirteenth the old token price.
The note ties the price move to GPT-5.6 serving optimizations, including self-tuned kernels, speculative decoding, KV caching, and leaner agent harnesses. Luna fell 80% to $0.20 per million input tokens and $1.20 per million output tokens, while Terra fell 20% to $2/$12. Sol also gained a Fast API mode, billed at 2x the price for up to 2.5x speed with the same stated intelligence. Latent Space cautions that benchmark-driven comparisons should be discounted, but says OpenAI is hard to beat on cost-effective, non-finetuned intelligence right now. Latent Space's note
The note ties the price move to GPT-5.6 serving optimizations, including self-tuned kernels, speculative decoding, KV caching, and leaner agent harnesses. Luna fell 80% to $0.20 per million input tokens and $1.20 per million output tokens, while Terra fell 20% to $2/$12. Sol also gained a Fast API mode, billed at 2x the price for up to 2.5x speed with the same stated intelligence. Latent Space cautions that benchmark-driven comparisons should be discounted, but says OpenAI is hard to beat on cost-effective, non-finetuned intelligence right now. Latent Space's note
score 7