Event date · · ByteDance

Reportedly: OpenAI releases GPT-6 Sol and Luna with 50% lower API prices

ByteDance 字节跳动Chinese AIGlobal API
Model Release

Reported by GeekPark · not yet confirmed by the company or a second independent outlet. We update this page when it is.

REPORTED

OpenAI released GPT-6 Sol and GPT-6 Luna on September 23, 2026, with API prices 50% lower than the GPT-5.6 promotional price, GeekPark reported. GPT-6 Sol outperformed Claude Opus 5 on AutomationBench at 9% of the cost per task.

China context

Outside China
Global API · geekpark.net
Claims
Company-reported; not yet independently evaluated
For builders
Developers outside China can access GPT-6 Sol and Luna via OpenAI's API at 50% lower prices than GPT-5.6 promotional pricing, with improved prompt caching for agent workloads.
For investors
OpenAI's price cut and performance gains on AutomationBench may pressure competitors like Anthropic to respond, potentially affecting market share and pricing power in the enterprise AI segment.

Translated from Chinese. Quotes and facts link to the original sources.

What happened

OpenAI launched GPT-6 Sol and GPT-6 Luna, two faster and more economical models trained using methods similar to GPT-6 Astra. The models bring Astra's performance in professional work, factuality, coding, computer use, and alignment to lower-cost tiers. API prices are 50% lower than the GPT-5.6 promotional price. In AutomationBench cross-application business process tests, GPT-6 Sol outperformed Claude Opus 5 at xhigh intensity while costing only 9% of Opus 5's per-task cost. GPT-6 Luna improved 5.4% over its predecessor at high intensity with a 58% reduction in per-task cost. On the Last Exam agent evaluation, GPT-6 Sol scored 56.4% at max effort, higher than Claude Opus 5's highest score, with 60% lower per-task cost. OpenAI also improved prompt caching for GPT-6, providing higher cache hit rates by default for faster agent responses.

Technical significance

GPT-6 Sol and Luna are distilled or otherwise optimized variants of GPT-6 Astra, retaining its capabilities in professional tasks, factuality, coding, computer use, and alignment while reducing inference cost. The models show strong performance on AutomationBench and Last Exam, with Sol matching or exceeding Claude Opus 5 at a fraction of the cost. Improved prompt caching increases cache hit rates, reducing latency and cost for agentic workloads that reuse context.

Industry impact

Developers using OpenAI's API see a 50% price reduction for GPT-6 Sol and Luna compared to GPT-5.6 promotional pricing, and agent builders can cut per-task costs by up to 91% versus Claude Opus 5 on AutomationBench. This pressures competing model providers to lower prices or differentiate on quality.

Decision value

The price reduction and improved caching lower the cost of deploying GPT-6-based agents, making advanced AI more accessible for businesses automating cross-application workflows. The performance-to-cost ratio on AutomationBench suggests significant savings for enterprises using agentic automation.

What to watch

A verifiable next signal is whether OpenAI extends the 50% price reduction beyond the promotional period or applies it to GPT-6 Astra. Another signal is whether Anthropic responds with price cuts or new model releases for Claude Opus 5.

Latest in Chinese AI

  1. DeepSeekReportedly: OpenDesign team's OpenMotion lets DeepSeek Harness create motion graphics
  2. XiaomiReportedly: Xiaomi adds MiMo UltraSpeed to Token Plan Pro and Max for a limited time
  3. MiniMaxMiniMax publishes cryptanalysis findings from its internal AI model
  4. DeepSeekReportedly: Simplex AI founder Tony Zhang open-sources Linger, a personal agent that can run locally
  5. MiniMaxMiniMax publishes Cyber repository on GitHub

All China AI Events

AIGC Newsletter

China AI, with sources and context.

Analysis of Chinese AI models, companies and policy, and what you can use outside China.