OpenAI Halves GPT-6 Pricing

OpenAI released GPT-6 Sol and GPT-6 Luna on Tuesday at 18:12 UTC, less than two hours after Anthropic shipped Claude Opus 5.5 — and only ten days after the chief executives of both companies publicly called for a slowdown in frontier AI development. Rather than pushing past its flagship GPT-6 Astra, OpenAI halved developer prices relative to GPT-5.6 promotional rates: Sol now costs $2 per million input tokens and $10 per million output tokens, while Luna lists at $0.10 and $0.50 respectively. Because API access is billed per token, the cuts pass directly through to the running costs of products built on top of the models. According to the launch announcement, both models build on the advances behind Astra to bring faster, cheaper AI to work at scale. The move lands directly in the orbit of Sam Altman, who co-founded Worldcoin (WLD) and still fronts the identity project.

OpenAI announced the launch in a post on X: https://x.com/OpenAI/status/2102460975790137662?ref_src=twsrc%5Etfw

Anthropic Answers With Opus 5.5

Anthropic moved first, releasing Claude Opus 5.5 at 16:31 UTC. The company says the model performs at the level of Claude Fable 5.1 on most tasks while costing 40% less to run than Opus 5, a figure drawn from its own tests on typical workloads. List prices fell 20%, to $4 per million input tokens and $20 per million output tokens.

Opus 5.5 is Anthropic's first release since CEO Dario Amodei published a September 12 essay urging laboratories to pace their capability gains — though he wrote that pacing “does not mean halting model training.” Outside reviewers METR and Frontier Design audited the model before launch, and Elon Musk, who had endorsed the slowdown call, publicly congratulated Anthropic on the release.

A Benchmark Arms Race

The price cuts are backed by benchmark claims. On AutomationBench, a Zapier-built test scoring whether an AI agent can complete full business workflows across 47 tools, GPT-6 Sol at its highest reasoning setting scored 33.2% at $0.27 per task — beating Claude Opus 5's 26.9% at more than 11 times the cost, according to OpenAI's own numbers. On Agents' Last Exam, Sol posted 56.4%, which OpenAI says tops Opus 5's best score at 60% lower cost per task. On OSWorld 2.0 computer-use tests, Sol nearly tied Opus 5, 60.5% to 60.3%, at roughly 80% less cost. The agentic focus matters for the pricing math: a single agent workflow can require many model calls, so per-token reductions scale up at the task level.

Alignment metrics also improved: coding-deception rates fell to 1.3% for Sol and 2.8% for Luna, down from 10.4% and 9.5% for GPT-5.6.

Standards, Not Licenses

Hours earlier the same day, OpenAI published a policy paper titled “Building standards for the next phase of AI,” proposing that the United States lead the drafting of global technical standards for frontier models through the emerging network of national AI safety institutes — bodies already established in Australia, Canada, Germany, France, Kenya, Japan, Korea, Singapore, India and the United Kingdom The proposal addresses recursive self-improvement, the point at which AI systems begin developing successor AI, and sets out three pillars: measuring RSI-related progress, defining human-oversight thresholds for automated research, and shared incident reporting. OpenAI stressed that such standards would not function as licenses or pre-market approvals. As a proposal, the paper's pillars would still need adoption by the institutes and their member governments before shaping any oversight practice.

Altman argued on X that people outside AI laboratories deserve a real say — and that standards should help prevent power concentration: https://x.com/sama/status/2102414347364335917

Slowdown Pledge Meets Skepticism

Skeptics never bought the slowdown pledge. Investor Michael Burry dismissed it as self-serving hype tied to Anthropic's expected IPO, while Morgan Stanley — an underwriter of that offering — said the push changes nothing for AI spending.

That skepticism has aged quickly: ten days after the CEOs' joint appeal, both laboratories shipped new models the same afternoon, and both cut prices rather than raising capability ceilings. Neither OpenAI nor Anthropic has said when a model above its current flagship will ship. To many market participants, the episode has read less as a slowdown than as the opening round of a price war.

Worldcoin (WLD) Traders Watch Altman's Twin Track

The arc of the day — same-afternoon launches, halved API prices and a governance blueprint — points to a frontier AI sector racing on cost while courting regulators at the same time. For Worldcoin, the signal matters because Altman's twin roles keep the token structurally tied to OpenAI's fortunes, and WLD's market capitalization has historically tracked OpenAI headlines more closely than its own product cycle. The linkage is largely a function of OpenAI's private status: with no OpenAI shares listed, WLD is one of the few market tickers tied to the company. The token has been treated as a listed proxy for sentiment around Altman's ventures — a pattern reflected in earlier coverage of his nine-member mathematics advisory panel, Karp's claim that OpenAI will never go public, and the projected $278 billion cash burn.

The primary record here is OpenAI's own launch post, which states that Sol and Luna build on the advances behind GPT-6 Astra — the fact on which WLD proxy trades rest.