OpenAI launched GPT-6 Astra, its latest flagship AI model, earlier this month. Now, the company is bringing those same architectural upgrades to its everyday engines. OpenAI officially released GPT-6 Sol and GPT-6 Luna, bringing Astra-level capabilities down to mid-tier prices with a massive 50% price cut for developers.
The announcement landed in the middle of a heated rivalry where Anthropic also dropped its new Opus 5.5 model.
Cutting prices in half while doubling factual accuracy
Instead of pushing raw model size, OpenAI focused on squeezing maximum performance out of fewer compute resources. According to the firm, GPT-6 Sol cuts factual error rates in half compared to GPT-5.6 Sol. This makes it a reliable senior staffer for heavy coding and computer-use automation.
Meanwhile, the lightweight GPT-6 Luna punches far above its weight. On high effort settings, Luna matches the capabilities of the previous generation’s Sol model at roughly one-hundredth of the cost.
The price drops are impressive across the board:
- GPT-6 Sol: Costs $2 per million input tokens and $10 per million output tokens, down 50% from promotional rates.
- GPT-6 Luna: Priced at just $0.10 input and $0.50 output per million tokens.
- Prompt Caching: Delivers up to 90% discounts on repeated inputs, making long-running AI agents dramatically cheaper to run.
Benchmark battles, safety guardrails, and cleaner prose
The release makes intriguing direct comparisons to Anthropic’s ecosystem. On benchmarks like AutomationBench and DeepSWE 1.1, GPT-6 Sol matched or approached Anthropic’s Fable 5 and Opus 5 models at up to 80% to 90% lower task costs. However, on OSWorld tests, GPT-6 Sol required maximum effort to match Opus 5 running at medium effort. This proves that OpenAI’s core pitch here is affordability over raw dominance.
On the safety front, alignment saw noticeable progress. In tests measuring whether models follow unauthorized instructions on a simulated message board, GPT-6 Sol took the unauthorized bait in only 11% of cases, down from 52% in GPT-5.6 Sol. The rate of restriction evasion also dropped, though flagship models like Astra still face scrutiny for attempting to work around human monitoring.
Both new models also drop corporate filler. They are adopting a revised communication style designed to provide shorter, direct responses with less jargon and higher factual precision.
Where you can try the new models today
OpenAI is rolling out the new tier across its entire app ecosystem. GPT-6 Sol and Luna are live for Plus, Pro, Business, Enterprise, and Edu subscribers in ChatGPT Work and Codex.
For Free and Go accounts, GPT-6 Luna is available in the desktop application as a fast, low-cost option for everyday queries and document summaries.
The post OpenAI Launches GPT-6 Sol and Luna: Double the Accuracy at Half the Price appeared first on Android Headlines.
​Â
