Published · 5h agoInvest2 min read
OpenAI prices GPT-6 Sol 50% below a promotional rate it is already running
OpenAI set GPT-6 Sol's API price 50% under a promotion it is already running on GPT-5.6, hours after Anthropic said Opus 5.5 costs about 40% less to run than Opus 5. Buyers on Microsoft's $30 seats and $200 credit packs see none of it.
Context for builders, not their beat.See today for builders
What happened
- Anthropic's Claude Opus 5.5 performs at the level of its flagship Fable 5.1, the company says, and costs around 40% less to run than the Opus 5 model it released in July.
- OpenAI set the API price for its new GPT-6 Sol and Luna models 50% below a promotion it is currently running on GPT-5.6, the predecessor to flagship GPT-6 Astra.
- Anthropic called Opus 5.5 the first release in a new family, with Sonnet 5.5 and Haiku 5.5 expected "over the coming weeks".
- Fortune reported that OpenAI's release undercuts Anthropic's on API pricing, based on each company's own announcement.
Compiled by The InvestorSomething wrong?How this is made
Why it matters
- constraint A halved price needs twice the billed tokens to stand still, so both labs' revenue now rests on usage growth their customers control.
- exposure Enterprises metered per seat and per credit at 0.8 cents get no pass-through until Microsoft moves its own prices; the discount reaches API customers only.
- decision Finance teams drafting 2027 plans can rebid the same workloads at a lower unit price, and that saving becomes the labs' growth only if the buyers spend it on more volume.
- contradiction Both labs publicly asked for a slowdown in September and then shipped cheaper models within hours of each other, so "pacing the frontier" is a restraint on capability alone.
The 50% cut OpenAI put on GPT-6 Sol is measured against a promotion. The base it picked is the promotion it is already running on GPT-5.6 [4]. Fortune published the two labs' per-token numbers in a table only, side by side [5].
Neither release is a major step up in capability, by Fortune's account [12]. OpenAI credits "improvements in caching and inference" for the reductions and says it is "passing those savings directly onto users and customers" [9].
Halve a price and the meter has to carry twice the tokens to hold revenue flat [1]. Anthropic's number is a cost number: about 40% less to run than the Opus 5 model it shipped in July, which makes Opus 5 roughly 1.67 times as expensive per unit of work [1][2]. Anthropic said it is "passing these efficiency savings on to our customers in the form of price cuts and rate limit increases" [2].
"OpenAI and Anthropic are engaged in a price war that is driving down the price of AI and therefore driving down their ability to profit from it and grow the price of models," Ara Kharazian, lead economist at Ramp, told Fortune [6]. "AI bulls assume that there will be highly performant models that provide more and more value, and therefore they should be more expensive" [7], he said. "But that is not how normal technology makes it to market" [8].
The cut lands on the API meter. Microsoft sells Microsoft 365 Copilot at $30 a user a month paid yearly [13] and Copilot Studio capacity in packs of 25,000 credits at $200 a month, which works out to 0.8 cents a credit [14][3]. Those are seat and credit prices, so a cheaper token reaches that buyer only when Microsoft reprices the pack.
"CFOs have seen some of the sticker shock, and they haven't seen some of the gains that were promised in the initial investments," Randall Hunt, CTO at the AI consulting company Caylent, said [15]. Caylent times how long employees take to finish tasks with AI tools and then computes a cost per task, an approach OpenAI encouraged in a mid-July 2026 blog post [16]. If buyers hold their task budgets and simply let the cheaper tokens shrink the bill, the labs absorb the cuts in their own revenue. The test is whether billed tokens more than double from here.
What to watch
- Whether OpenAI's GPT-5.6 promotion becomes standing list pricing, which would fix the base the 50% cut is measured from.
- Whether Microsoft reprices the $200 credit pack or the $30 Copilot seat now that underlying token prices have fallen.
- Whether Anthropic's Sonnet 5.5 and Haiku 5.5 land below Opus 5.5 on price per token when they ship.
Claim ledger
Ranked by verification strength, evidence, and original report placement.
- [1]
Anthropic debuted Claude Opus 5.5, which it says performs at the level of its flagship Fable 5.1 model but costs around 40% less to run than Opus 5, which came out in July.
- [2]
Anthropic said the model is its most efficient, and it is "passing these efficiency savings on to our customers in the form of price cuts and rate limit increases."
- [3]
OpenAI released GPT-6 Sol and GPT-6 Luna, offshoots of its flagship GPT-6 Astra model that came out earlier this month, intended to offer a version of Astra for everyday work.
ReportedView cited source - [4]
OpenAI slashed the API cost by 50% lower than the pricing on a promotion it is currently running for Astra's predecessor, GPT-5.6.
ReportedView cited source - [5]
Fortune reported that OpenAI's release offers lower API pricing than Anthropic's, based on each company's announcement, as shown in a table accompanying the article.
ReportedView cited source - [6]
"OpenAI and Anthropic are engaged in a price war that is driving down the price of AI and therefore driving down their ability to profit from it and grow the price of models," Ara Kharazian, lead economist at Ramp, told Fortune.
- [7]
"AI bulls assume that there will be highly performant models that provide more and more value, and therefore they should be more expensive," Kharazian said.
- [8]
"But that is not how normal technology makes it to market," Kharazian said.
- [9]
OpenAI attributes the cost reductions to "improvements in caching and inference" and says it is "passing those savings directly onto users and customers."
- [11]
These are the first model releases since both companies publicly called to "pace the frontier": Anthropic CEO Dario Amodei published a blog post laying out a slowdown plan in mid-September, which OpenAI CEO Sam Altman supported in an X post.
ReportedView cited source - [12]
Fortune reported that the releases from both companies are not frontier models, in that they are not a major step up in capabilities.
- [13]
Microsoft lists Microsoft 365 Copilot at $30.00 per user per month, paid yearly.
ReportedView cited source - [14]
Copilot Studio is sold as a tenant-wide license that includes Copilot Credit capacity packs of 25,000 Copilot Credits each, priced at $200.00 per pack per month.
ReportedView cited source - [15]
"CFOs have seen some of the sticker shock, and they haven't seen some of the gains that were promised in the initial investments," Randall Hunt, CTO at the AI consulting company Caylent, said.
- [16]
Caylent calculates how long it takes employees to complete tasks with AI tools and then calculates "cost per task," an approach OpenAI encouraged in a mid-July 2026 blog post.
ReportedView cited source - [18]
Anthropic said Opus 5.5 is the first release in a new family of models, with Sonnet 5.5 and Haiku 5.5 expected "over the coming weeks."
ReportedView cited source - [d1]
Holding revenue flat after a 50% price cut requires billed token volume to double.
Derived - [d2]
A 40% reduction in cost to run implies Opus 5 was about 1.67 times as costly per unit of work as Opus 5.5.
Derived - [d3]
Microsoft's Copilot Credit capacity pack works out to $0.008, or 0.8 cents, per credit.
Derived
Sources & coverage · 2 publishers
The reporting this story was synthesized from, earliest first. Every link goes to the original.
Additional citations
- Fortune, reporting Anthropic's announcement
- Anthropic, via Fortune
- Ara Kharazian, lead economist at Ramp, to Fortune
- Ara Kharazian, Ramp
- OpenAI, via Fortune
- Fortune
- Randall Hunt, CTO, Caylent, via Fortune