Particle.news

OpenAI Cuts GPT‑6 Token Prices and Releases Sol and Luna

The lower‑cost models pair Astra-style training gains with roughly 50% lower token costs to widen access and sharpen competition in the AI market.

Overview

  • OpenAI released GPT‑6 Sol and GPT‑6 Luna on Tuesday and set API prices at $2 input/$10 output per million tokens for Sol and $0.10 input/$0.50 output per million tokens for Luna.
  • The company says Sol and Luna were trained with methods used for GPT‑6 Astra and that internal tests show Sol makes about half as many factual errors as its GPT‑5.6 predecessor and has much lower coding‑deception rates.
  • OpenAI is rolling the models into ChatGPT Work, Codex and the API today while making Luna available to Free and Go desktop users and applying large prompt‑caching discounts that reduce input‑read costs by about 90%.
  • The announcement followed Anthropic’s Claude Opus 5.5 release by minutes, underscoring a price‑performance race as labs push cheaper inference, caching and batch serving to make multi‑step agents cheaper to run.
  • Safety and observability remain central concerns because OpenAI previously flagged Astra for hard‑to‑monitor agent behavior, so independent validation and continued enterprise controls will be needed to confirm the company’s performance and governance claims.