GPT-6 gets cheaper, and Claude cuts costs too
OpenAI's new GPT-6 Sol and Luna models offer lower API prices, and updated prompt caching could cut costs further for apps that repeat the same instructions. Anthropic has a faster Claude Opus 5.5, while Meta's Muse raises a practical question: what access should we give an AI assistant?
The Big Story
OpenAI News
Introducing GPT-6 Sol and Luna
OpenAI launched GPT-6 Sol and Luna, lower-cost models for demanding work and routine, high-volume tasks, respectively. Its published API rates per million tokens are $2 input and $10 output for Sol, and $0.10 input and $0.50 output for Luna (down from GPT-5.6 promotional rates of $4/$20 and $0.20/$1.20). OpenAI reports fewer factual mistakes for Sol in an internal test, but says those error-prone test conversations aren't representative of typical use. Both models are available through the API and ChatGPT Work and Codex on paid plans, with Luna accessible to Free and Go users in the desktop app and a broader ChatGPT rollout planned.
For repetitive API tasks, compare Luna and Sol against current models using real examples and total token costs; ChatGPT subscribers can try them as access appears.
For Builders
OpenAI News
Better prompt caching for GPT-6
OpenAI updated GPT-6 prompt caching so repeated API requests can reuse shared instructions and context more often. Eligible prompt prefixes reused within 30 minutes can receive discounts of up to 90% on cached input tokens. Developers also get a cache-hit dashboard, diagnostics for misses, explicit cache breakpoints, and a way to adjust reasoning effort without invalidating cached context.
This matters if a business runs an AI app through OpenAI's API, not simply uses ChatGPT. Check cache-hit rates and costs before changing prompts or tool definitions that the app reuses.
Anthropic News
Introducing Claude Opus 5.5
Anthropic released Claude Opus 5.5, saying it performs at the level of Claude Fable 5.1 on most work while costing 40% less per typical workload than Opus 5. API pricing is $4 per million input tokens, $20 per million output tokens, and $0.20 per million cache-read tokens; Anthropic also says output is more than 30% faster. Five-hour limits are increasing for Pro, Max, Team, and seat-based Enterprise subscribers, while access for some biology and cybersecurity uses requires verification.
If your team pays for Claude API usage, compare real task costs and quality against Opus 5 before switching; the lower token price alone doesn't tell the whole story.
For Business
TechCrunch AI
Meta admits Muse’s likeness to OpenClaw isn’t a coincidence
Meta product chief Nat Friedman said Muse, its AI assistant, was heavily inspired by the open-source OpenClaw agent but built from scratch. Users had spotted matching workspace file names and reportedly near-identical content in a file defining the assistant's behavior; Friedman didn't dispute those comparisons. Muse recently reached No. 1 on the U.S. App Store, but Meta offered no further detail about the overlap.
Similar-looking agents aren't necessarily interchangeable. Before letting either access business accounts, compare their permissions, data handling, and approval controls on a small, reversible task.