What Slowdown? OpenAI & Anthropic Cut Prices With Sol, Luna, and Opus 5.5
On Tuesday, Sept. 22, 2026, Anthropic and OpenAI dropped cheaper workhorse models within roughly ninety minutes of each other — Claude Opus 5.5 first, then GPT-6 Sol and GPT-6 Luna — turning mid-September “pace the frontier” talk into a same-day enterprise price duel. Coverage from TechCrunch, Ars Technica, Fortune, and The Register frames the twin launches as an affordability race, not another flagship demo day.
This is the first major dual-lab model wave since both companies’ mid-September calls to slow the frontier cadence. The product bet this time is cost-per-task: more capability for everyday enterprise spend, priced to win the stack that CFOs actually renew.
Anthropic: Claude Opus 5.5
Per Anthropic’s Opus 5.5 announcement, Claude Opus 5.5 is the first model in the Claude 5.5 family. Anthropic says it delivers Fable 5.1–level performance on most work at roughly 40% lower typical-workload cost versus Opus 5. API pricing is listed at $4 per million input tokens and $20 per million output tokens — about 20% less than Opus 5 — with cache reads at $0.20 per million tokens (about 60% less). Anthropic also claims more than 30% faster output.
The company positions Opus 5.5 as its first release since CEO Dario Amodei’s “pace the frontier” call. Anthropic cites external evaluators including Frontier Design and METR, calls out its strongest automated behavioral audit score to date, and says cyber and biology safeguards are similar to Fable 5.1. Availability spans the Claude Platform (claude-opus-5-5), AWS, Google Cloud, and Azure. Sonnet 5.5 and Haiku 5.5 are described as “coming weeks.”
Those performance, cost, and audit claims are Anthropic’s. Independent third-party leaderboards beyond what cited outlets report are not invented here.
OpenAI: GPT-6 Sol and GPT-6 Luna
About ninety minutes later, OpenAI expanded the GPT-6 lineup after Astra with GPT-6 Sol and GPT-6 Luna, per its community announcement. OpenAI says promotional API pricing for Sol and Luna runs 50% lower versus GPT-5.6 Sol/Luna, with caching and inference discounts in the mix. Sol is aimed at complex coding and professional work; Luna targets high-volume clerical tasks.
Both models land in the API. ChatGPT Work and Codex access cover Plus, Pro, Business, Enterprise, and Edu tiers. Luna also ships on Free and Go desktop. TechCrunch and Ars Technica report Sol API pricing around $2 / $10 per million input/output tokens and Luna around $0.10 / $0.50 — figures attributed to those outlets’ coverage of the launch, not independent JamoraquAI metering.
Price War Framing: Fortune, Ars, Register
Fortune casts the twin drops as an enterprise price war despite recent slowdown solidarity, citing a Ramp economist on profit pressure from cheaper models and outright cuts. Ars Technica’s same-day read is blunt: both labs promise a little more for a lot less money. The Register’s Wednesday, Sept. 23 piece adds cadence context — Opus 5.5 roughly 21 days after Fable/Mythos 5.1, Sol/Luna about 19 days after GPT-6 Astra (Sept. 3) — and notes Artificial Analysis putting Opus 5.5 on par with Astra for agentic work on the Intelligence Index.
That Register citation is the only comparative ranking used here. Lab self-reported benchmarks stay attributed to Anthropic and OpenAI; the “pace the frontier” irony is editorial framing, not a claim that either CEO contradicted a binding pledge.
JamoraquAI Take
When the same labs that just asked the industry to pace the frontier ship cheaper workhorse models within hours of each other, the real race isn’t who’s smartest — it’s who owns everyday enterprise spend when CFOs demand cost-per-task, not another flagship demo.
Flagship launches still move markets. Procurement moves on unit economics. Opus 5.5’s typical-workload cost cut, Sol’s coding-tier promo pricing, and Luna’s clerical-volume lane are not subtle signals: the battleground shifted from demo-reel intelligence to the invoice line that renews every quarter.
For buyers, the near-term question is which workhorse clears the bar on quality at the price CFOs will actually approve — and which lab locks that spend before Sonnet/Haiku 5.5 and the next GPT-6 sibling land.
Sources
- Anthropic — Claude Opus 5.5 announcement, Sept. 22, 2026
- OpenAI Community — Announcing GPT-6 Sol and GPT-6 Luna in the API, Codex, and ChatGPT, Sept. 22, 2026
- TechCrunch — OpenAI launches GPT-6 Sol and Luna, Sept. 22, 2026
- Ars Technica — New Anthropic, OpenAI models make same promise: a little more for a lot less money, Sept. 22, 2026
- Fortune — What AI slowdown? OpenAI, Anthropic release dueling more-affordable models as AI price wars heat up, Sept. 22, 2026
- The Register — Frontier AI keeps racing despite calls to slow down, Sept. 23, 2026
