Anthropic and OpenAI Ship Cheaper Models 101 Minutes Apart, One Week After "Pace the Frontier" Pledge

On September 22, Anthropic and OpenAI released cheaper new models just 101 minutes apart, which ThursdAI's podcast treated as the core AI story of the week. The timing is especially sensitive: on September 12 Dario published an essay urging labs to "pace the frontier," signed by Sam Altman, Elon Musk, Demis and other lab leaders. Barely a week later, the two labs shipped stronger, cheaper models almost simultaneously, leading commentators like @altryne to declare the pledge effectively dead.

Confirmed

On September 22, Claude Opus 5.5 launched at 09:31 PT; OpenAI followed 101 minutes later, with both cutting prices
Per the ThursdAI podcast (hosts @altryne, @petergost), Opus 5.5 beat Fable 5.1 on GDPval, costs 40% less than Opus 5, and ranks first on Code Arena
Opus 5.5 claims a return to Opus 4.6-level intelligence with a more human tone; it even acted as a live producer on the ThursdAI stream this week
Per DL Weekly, OpenAI responded within hours, pricing GPT-6 at half the previous 5.6 tier and claiming GPT-6 (codename Sol) makes about half the factual errors of its predecessor; @altryne found GPT-6 Sol better value than his own daily driver
Meta betting on the Muse model was also discussed on the show
The next ThursdAI episode goes live October 1, 11am PT from Moscone Center in San Francisco at a CoreWeave event

Why it matters

Labs verbally committed to coordinated slowdowns yet raced neck-and-neck on the same day with a price war, exposing the tension between pledges and commercial competition
Cheaper and stronger models will further intensify price and capability competition, with clear cost-structure implications for the downstream application ecosystem

2026-09-25 ~ 2026-09-27 · 13 related posts

Primary sources

101 minutes apart, Anthropic and OpenAI break the 'pace the frontier' pledge with cheaper models — altryne · 2026-09-25 14:17 (UTC+08:00)
OpenAI prices GPT-6 at half the 5.6 tier, claims Sol halves factual errors — dl_weekly · 2026-09-27 01:00 (UTC+08:00)
Podcast: Opus 5.5 Beats Fable 5.1 on GDPval at 40% Lower Cost, Tops Code Arena — thursdai_pod · 2026-09-25 23:20 (UTC+08:00)

[source] 101 minutes apart, Anthropic and OpenAI break the 'pace the frontier' pledge with cheaper models — altryne · 2026-09-25
Anthropic and OpenAI Ship Cheaper Models 101 Minutes Apart, Igniting a Price War — altryne · 2026-09-25
[source] Podcast: Opus 5.5 Beats Fable 5.1 on GDPval at 40% Lower Cost, Tops Code Arena — thursdai_pod · 2026-09-25
ThursdAI to broadcast live from CoreWeave's FullyConnected26 at Moscone — thursdai_pod · 2026-09-26
Opus 5.5 returns 40% cheaper, OpenAI ships GPT 6 Sol and Terra, Meta goes all in on Muse — altryne · 2026-09-26
Anthropic and OpenAI ship cheaper models 101 minutes apart; GPT-6 Sol beats Astra on price — altryne · 2026-09-27
[source] OpenAI prices GPT-6 at half the 5.6 tier, claims Sol halves factual errors — dl_weekly · 2026-09-27

6 near-duplicate retellings: thursdai_pod · altryne · thursdai_pod · altryne · thursdai_pod · thursdai_pod