All news

Frontier AI prices just fell again — Claude Opus 5.5 and GPT-6 Sol/Luna cut rates in hours

Anthropic cut Claude Opus 5.5 by 20% to $4/$20 per M tokens; OpenAI’s GPT-6 Sol and Luna land at roughly half GPT-5.6. What UK SMBs should check.

23 September 2026 4 min read
Frontier AI prices just fell again — Claude Opus 5.5 and GPT-6 Sol/Luna cut rates in hours

By Jason Sibley

What happened

On 22 September 2026 Anthropic shipped Claude Opus 5.5 with a 20% price cut, taking it to $4 per million input tokens and $20 per million output tokens. Roughly an hour later OpenAI released GPT-6 Sol and GPT-6 Luna at about half the price of their GPT-5.6 predecessors. That landed on the heels of Grok 4.7 and Xiaomi’s MiMo v2.6 the day before — four notable model updates in about 48 hours.

Reported API list prices include GPT-6 Luna at $0.10/$0.50 per million tokens (input/output) and GPT-6 Sol at $2/$10, with Grok 4.7 at $2/$6. Full breakdown and timeline: TPS Report.

What it means for a small business

Cost. If you pay for API usage (chatbots, document tools, internal assistants), the mid-tier cuts matter more than flagship marketing. Luna in particular is priced for volume work; Opus 5.5 is cheaper than earlier Opus releases but still a premium tier.

Time. Switching models mid-project burns hours on prompt re-testing. Cheaper does not mean drop-in identical. Plan a short bake-off before you move production traffic.

Risk. New model, new quirks. Early testers noted Opus 5.5 can burn through long reasoning budgets on simple tasks. Cap output tokens and keep a human check on anything customer-facing or money-related.

Vendor lock-in. With OpenAI, Anthropic, xAI and others racing on price, it is easier to justify a second provider so you are not stuck if one raises rates or throttles.

What you can do this week

  1. List where you actually spend. Note which tools bill by token or seat, and which models they sit on today.
  2. Run a three-prompt bake-off. Same three real jobs (a client email, a quote summary, an internal FAQ answer) on your current model and one cheaper alternative; score accuracy and tone, not just price.
  3. Set a spend alert. Cap monthly API or workspace AI spend so a busy week cannot surprise you.
  4. Keep a fallback. Document which secondary model or tool you would switch to if the primary one changes price or behaviour overnight.

How we can help

If you want a plain-English map of which model tier fits your workload, without chasing every launch, our free website review, AI sales & marketing desk, and SMB AI sessions are built for that kind of practical choice.

Get the full step-by-step guides

Create a free account for the complete playbooks, walkthroughs and frameworks — everything you need to actually put AI to work. No card, no cost.

My AI Apprentice

Helping non-technical adults use AI on their own terms. Imprint partner: Foundry Works Press.

Stay in the loop

Not ready to join yet? Drop your email and I'll keep you posted.

© 2026 My AI Apprentice / Foundry Works Press. All rights reserved.

Built with intention, not hype.