Skip to content
AI Metric

Chris M.

The cheap model won on the work that pays

Most AI business cases in construction were priced on one assumption: the good model is dear, so you use it sparingly and put up with the cheap one everywhere else. That assumption stopped holding this week, and it stopped holding on the vendor's own figures rather than ours.

Written up on 28 September, covering the week of 22 September. Every figure below was read at the publisher on 28 September.

Half the price, and it came top

What shipped. OpenAI released GPT-6 Sol and GPT-6 Luna on 22 September, cutting the price of both by half against what the previous generation charged. Sol went from four dollars to two per million input tokens, Luna from twenty cents to ten. On AutomationBench, a test of business workflows across 47 tools, Sol scored 33.2 per cent at 27 cents a task. Claude Opus 5 scored 26.9 per cent at 11.1 times that cost.

Where it lands. The cheap tier is now the right home for most of what a contractor would actually automate: reading a delivery ticket, pulling dates off certificates, turning a week of records into a draft narrative. That work is high volume and checkable, which is exactly the shape the cheap tier handles. Anyone still paying flagship rates to summarise documents is paying for reasoning they are not using.

Monday morning. Take one process you priced on last quarter's rates and price it again on this week's. The cheapest tier is a tenth of what it cost in July.

OpenAIGPT-6 API pricing
OpenAI's published pricing table for GPT-6 Sol and Luna, showing GPT-5.6 Sol at four dollars input and twenty dollars output falling to two dollars and ten dollars for GPT-6 Sol, and GPT-5.6 Luna at twenty cents and one dollar twenty falling to ten cents and fifty cents for GPT-6 Luna, each marked fifty per cent cheaper, per one million tokens.
The whole of the price change, on the publisher's own page. Fifty per cent off both tiers, stated plainly, with the reason given as caching and inference improvements passed on.

Anthropic cut its own price the same day

What shipped. Claude Opus 5.5 landed on 22 September at four dollars input and twenty dollars output, about 40 per cent cheaper on typical workloads than Opus 5 and generating output more than 30 per cent faster. Reading a cached token fell from fifty cents per million to twenty. Anthropic report a tester completing a 680,000 line code migration in under a day.

Where it lands. Two vendors cutting price on the same day is a market telling you something: the margin has moved from the model to what you do with it. For a business buying AI, that means a contract signed on today's rates is a worse deal in ninety days, and the tools should be swappable rather than welded in.

Monday morning. Check whether anything you are being sold this quarter locks you to one model. If it does, ask what happens at the next price cut, because there will be one.

AUTOMATIONBENCH 1.0.6, COST PER TASK AGAINST SCOREbusiness workflows across 47 tools, in sales, operations, finance, support, marketing and HRGPT-6 Solxhigh$0.2733.2%Claude Fable 5.1max, with Opus 5 fallbackover $2.4031.4%GPT-6 Astralow$1.0530.3%Claude Opus 5max$3.0026.9%SCOREHOW TO READ THISEach bar is what one finished task cost. The figure on the right is the score that money bought.The shortest bar carries the highest score, which is the whole of the point.Only GPT-6 Sol's 27 cents is published directly. The other three are OpenAI's own multiples of it.The dashed bar is open because its publisher says that figure leaves out the cost of its fallbacks.
The shortest bar carries the highest score. Price and capability used to move together, and on this test in this week they came apart.Source: OpenAI, GPT-6 Sol and Luna announcement, read 28 September 2026
ModelInputOutputWhere it earns its place
GPT-6 Astra$10$50The judgement call at the end, not the volume
Claude Opus 5.5$4$20Long agentic work, computer use, research
GPT-6 Sol$2$10Multistep work that gets checked anyway
GPT-6 Luna$0.10$0.50Reading, extracting, filing, at volume

Caching became the biggest lever on the bill

What shipped. Less visible than the price cuts and worth more. Cached input on GPT-6 is discounted by 90 per cent, and OpenAI report that GitHub cut the share of prompt tokens needing fresh processing by more than half across billions of requests. Opus 5.5 reads cached tokens at a fifth of what Opus 5 charged.

Where it lands. An agent does not answer once. It plans, looks things up, calls a tool, checks itself and goes round again, so the bill is the number of passes multiplied by what each pass burns. Most of that is the same unchanging context every time: the contract, the register, the standing instructions. Structured so it is reused, it is nearly free. Structured badly, you pay full price for it on every lap.

Monday morning. If you are running anything in production, ask whoever built it what your cache hit rate is. If nobody knows, that is the cheapest performance work available to you this month.

Matthew Bermanthe launch, followed
Where the OpenAI half of this week got noticed. The read is his; every figure above was taken off the publishers' own pages instead, which is the rule here. GPT-6 SOL AND LUNA ARE OUT!!!, published by Matthew Berman on YouTube.
Matt Wolfethe Anthropic half, followed
The other side of the same day. Worth watching for the working through of what the price change actually buys, rather than the benchmark table. Claude Opus 5.5 Didn’t Need to Go This Hard, published by Matt Wolfe on YouTube.

The question to ask before the next one ships

There will be another release in about six weeks, and the honest position is that the models are already ahead of most businesses' ability to use them. The limit is not the model. It is whether the work you would point it at is written down anywhere it can reach.

If you have a process eating hours every week and you suspect this month changed the maths on it, that is worth twenty minutes.

Sources

Every figure and price in this edition was read from the page linked below on 28 September 2026. Where a number could not be confirmed at its publisher, it is not in the edition.

Followed this week

Where the week was picked up. Every figure and price above was then checked at the publisher and drawn from that source, so no chart here is traced off a video.

This series reads the week’s AI releases from a construction delivery position, not a technology one. If one of these lands on a package you are running and you want to talk about what it would take, book a 30 minute call.

Commentary on publicly announced capability, written from twenty-plus years of Tier 1 delivery. It is not design guidance, fire or building safety advice, or contractual advice, and no live project is described.