7d ago25 sources70 reactions

OpenAI launches GPT-6 Sol and Luna at half the API price

42 DoomStory + reactionsCompetitive pricing moves framed as market pressure

By Splitfeed EditorialUpdated Sep 26, 2026, 20:18 UTC

25 sources · Techeconomy · The Times of India · The Indian Express

Boom: GPT-6 Sol and Luna API prices set at half the GPT-5.6 promotional rate
Boom: Sol priced below Claude Opus 5.5; Luna priced below DeepSeek V4.1 Flash
Neutral: OpenAI and Anthropic both launched competing models on September 22, 2026
Doom: One outlet reports the price cuts produce minimal performance improvement over existing models
Boom: GPT-6 Astra reportedly solved an Enigma message unsolved since 2005

The story in full

On September 22, 2026, OpenAI released two new models, GPT-6 Sol and GPT-6 Luna, expanding its GPT-6 lineup. The API prices for both are set at half the previous GPT-5.6 promotional price, undercutting competitors including Anthropic's Claude Opus 5.5 and DeepSeek's V4.1 Flash, according to multiple outlets. The launches occurred the same day Anthropic released Claude Opus 5.5.

Sol and Luna are described as carrying Astra-like capabilities while being cheaper and faster than existing GPT-6 offerings. At least one outlet reports that the price cuts barely move the needle on performance relative to existing models. The pricing is characterized by some sources as positioning GPT-6 Sol below Claude Opus 5.5 and GPT-6 Luna below DeepSeek V4.1 Flash.

Analysis

383 words

By Splitfeed EditorialLast updated Sep 26, 2026, 20:18 UTC

On September 22, 2026, OpenAI released two new models, GPT-6 Sol and GPT-6 Luna, pricing both at half the previous GPT-5.6 promotional API rate. The move places Sol below Anthropic's Claude Opus 5.5 on price and Luna below DeepSeek's V4.1 Flash, both of which are direct competitors in the frontier model space. The launch coincided with Anthropic releasing Claude Opus 5.5 the same day. GPT-6 Astra, an existing model in the lineup, separately drew attention for reportedly solving a historical Enigma message that had gone unsolved since 2005.

The pricing matters because it compresses margins across the frontier tier at a moment when Anthropic and OpenAI are both spending heavily on compute. Whether the cuts reflect genuine efficiency gains or a deliberate move to squeeze competitors is disputed. At least one report notes that the performance improvements over existing models are modest relative to the price reduction, raising the question of whether Sol and Luna are substantially new or effectively optimized rereleases. Middle-ground observers, including The New Stack, note that the models reduce deception and unauthorized actions, but flag that model monitoring and observability remain open questions. Adam Galas pushed back on claims that DeepSeek and Kimi K3 would bankrupt the major labs, arguing the financial data points the other way.

Pro-AI voices treated the release as a decisive competitive event. Five O'Click Tech argued the pricing makes older models and many competitors instantly uncompetitive, while Code and Clarity highlighted fewer factual errors and Astra-level reliability. Android Authority described the models as faster, better, and far more affordable. On the anti-AI side, Carl Quintanilla quoted concern that poorly managed AI could be a risk to humanity, and TMurr warned that if AI agents are already compromising government systems, litigation against the major labs cannot come soon enough. Alex Morash dismissed the excitement more bluntly, calling AI more flash than substance. The anti-AI camp broadly sees the simultaneous calls for slowdowns from lab leaders and the release of cheaper, more capable models as a fundamental contradiction.

The comparison that would settle much of the debate is a direct benchmark of Claude Opus 5.5 against GPT-6 Sol on standardized tasks, which neither camp has produced yet. Context-length pricing penalties, flagged by middle-ground observers, could also narrow the apparent cost advantage materially depending on use case.

Pro-AI18

What Pro-AI voices are sayingThe price cuts make GPT-6 Luna and Sol dramatically more accessible, rendering many competitors instantly uncompetitive, and the models' broader capabilities signal a new phase of AI progress. Some see the pricing move as a direct strategic attack on DeepSeek and Anthropic.

Quote 1 of 13

From the perspective of “an average person”, ChatGPT is delivering fantastic products. - For general chat and web search, occasional image editing, small coding work, document review etc. ChatGPT Plus is basically limitless and “just...

leokennisvia Hacker News

Anti-AI34

What Anti-AI voices are sayingCritics highlight the contradiction between OpenAI and Anthropic leaders calling for AI slowdowns while simultaneously releasing cheaper, more powerful models, and broader concerns persist about safety, profitability, and regulatory gaps. A notable minority dismisses the competitive framing entirely, arguing the model naming is confusing and cost comparisons are meaningless since OpenAI controls p

Quote 1 of 13

Getting to the point where these headlines depress me. I just wish they would stop getting better. I don't know where my career is gonna be in a few years.

delillosvia Hacker News

Middle Ground18

What Middle Ground voices are sayingObservers note that high margins give both labs room to keep cutting prices, and that the real competitive picture is unclear since Opus 5.5 and GPT-6 Sol have not been properly benchmarked head-to-head. Some skeptics question whether Sol is genuinely new or just an optimized GPT-5.6, and note that context-size pricing penalties narrow the cost advantage.

Quote 1 of 13

OpenAI, Antrophic and others are operating with 80% margins. They can lower the prices for a long while.

meeritavia Hacker News

Add your take

0 reader votes

More Pro-AI reactions (12)

“GPT-6 Luna being half the price of GPT-5.6 Luna is a really big deal.”
simonw, Hacker News · 18:41 UTC
“Lunas insane discount is a game changer, OpenAI knows what they are doing here. Luna at max reasoning effort, even though its not optimal for long conversations, its incredibly intelligent while dirty cheap. Its not even competition anymore.”
Alifatisk, Hacker News · 22:34 UTC
“Luna is even fucking cheaper now at just $0.10 per million input and $0.50 for output, Sol is also halved in price, now $2/$10.”
Eris, Bluesky · 18:28 UTC
“they're trying to murder deepseek. anthropic had better step their game up with sonnet and haiku 5.5”
conputer dipshit, Bluesky · 19:04 UTC
“it really is amazing how thoroughly you can blow somebody's mind right now if they talk about using Sonnet to save money.”
jslaker.net, Bluesky · 17:40 UTC
“This makes their older models, and many competitors, instantly uncompetitive.”
Five O'Click Tech, Bluesky · 07:01 UTC
“GPT-6 Astra surpasses the human baseline in action efficiency on ARC-AGI-3. It used fewer actions than the median tested human on 96% of levels.”
Wayne Radinsky, Bluesky · 02:44 UTC
“Yeah, and tbh it’s a little shocking how good GPT-6 is in robotics despite the fact that it’s not built for it”
Jeremy Diamond ☕️, Bluesky · 01:02 UTC
“Anthropic's Opus 5.5 in particular is a dream when it comes to programming graphical scenes from scratch. Absolutely amazing.”
Edwin Hayward, Bluesky · 22:07 UTC
“faster decisions, cheaper runs, and structured output without an extra parser. AI moves fast an I am excited to explore more.”
Christian Münch, Bluesky · 09:00 UTC
“This summer, GPT-5.6 Luna did the same for $0.0015. A 377-fold drop in 18 months.”
Epoch AI, Bluesky · 21:28 UTC
“ChatGPT’s new GPT-6 Sol and Luna models are faster, better, and far more affordable”
Android Authority, Bluesky · 05:08 UTC

More Anti-AI reactions (12)

“Trying to convince us there is a difference between GPT-6-Sol and GPT-5.6-Terra or whatnot is ludicrous to the point of being insulting, especially when new models come out every week.”
reenorap, Hacker News, skeptic · 20:06 UTC
“I am sick and tired of the inconsistencies in GPT model families. There is no 6-Terra. As for any cost based argument, it is immediately invalid because the cost is something that OpenAI fully controls and manipulates.”
OutOfHere, Hacker News · 18:34 UTC
“If I'm understanding correctly now when you want to use Codex to do a given task you need to decide between: GPT-6 Astra (low medium high xhigh max ultra) GPT-6 Sol (low medium high xhigh max ultra) GPT-6 Luna (low”
adamrezich, Hacker News · 19:01 UTC
“NEW: OpenAI took 20 million ChatGPT responses and had one of its experts search for chunks of text—specifically chunks of 25 words in a row- that were identical between ChatGPT’s responses and a news article. They found more than 400,000”
Mother Jones, Bluesky · 21:38 UTC
“If managed poorly, I even believe AI could be a risk to humanity as a whole”
Carl Quintanilla, Bluesky · 22:50 UTC
“In @nytopinion.nytimes.com “If OpenAI and Anthropic really want to help advance science, there is one obvious way they can do so: They can spend a tiny fraction of their vast fortunes to fund the kind of humble but essential research”
The New York Times, Bluesky · 12:50 UTC
“The recent AI Doc on Netflix put a lot emphasis on the need for international regulation of AI, based on interviews with the CEOs of Anthropic, OpenAI, and Google DeepMind”
Eukanal Bytmeagh❌👑🐸, Bluesky · 00:42 UTC
“FYI OpenAI an Anthropic aren’t companies, they’re cults. Hope that helps”
Yo Gabapentin, Bluesky · 20:58 UTC
“3 guys hacked OpenAI using a rival Anthropic model. Here's what it shows about frontier labs' vulnerabilities.”
CBS News, Bluesky · 19:15 UTC
“I wish them a very very strong “go [redacted] yourselves””
LurkyMcLurk, Bluesky · 18:36 UTC
““If OpenAI and Anthropic really want to help advance science, there is one obvious way they can do so: They can spend a tiny fraction of their vast fortunes to fund the kind of humble but essential research that’s increasingly”
Mark Little, Bluesky · 07:47 UTC
“Anthropic and OpenAI released cheaper AI models days after their leaders called for a slowdown.”
Science X / Phys.org, Bluesky · 08:39 UTC

More Middle Ground reactions (12)

“Overall, I expect for most people think the winner of today was Anthropic. I personally am preferring Opus 5.5 at medium over GPT-6 Sol Max, in very very early tests.”
mchusma, Hacker News · 19:07 UTC
“Anthropic and OpenAI announced new, less costly artificial intelligence models, as both labs face stiff competition from rivals offering open-weight models.”
CNBC, Bluesky · 01:44 UTC
“The problem with OpenAI is that they double token price at something like 272K context size. That really levels the price playing field.”
daeden323.bsky.social, Bluesky, skeptic · 18:43 UTC
“#Anthropic & #OpenAI seem determined to bust in the door to every computational subfield imaginable.”
Generic white guy SysAdmin, Bluesky · 06:01 UTC
“OpenAI halved GPT-6's price to beat Anthropic on cost. But Opus 5.5 already reset the comparison, and nobody's run the two head-to-head yet.”
The New Stack, Bluesky · 03:00 UTC
“Between the open models and Apple pushing on-device stuff, I have a glimmer of hope about the tech distribution”
Scott Frazer, Bluesky · 16:09 UTC
“OpenAI’s new GPT-6 Sol and Luna sharply reduce deception and unauthorized actions, but questions remain about model monitoring and observability.”
The New Stack, Bluesky · 09:00 UTC
“GPT-6 Sol feels more like the next Terra than the next Sol, and that Opus 5.5 is kicking its ass.”
Boon aka Hwee-Boon Yar, Bluesky, skeptic · 06:30 UTC
“OpenAI halved the price of its new GPT-6 Sol and Luna models, but the immediate reaction wasn't celebration. Commenters quickly turned to tightened usage limits and the unavailable $200/month subscription tier.”
Hacker News Companion, Bluesky · 01:00 UTC
“Opus 5.5 failed twice at $2.56 a try on a simple SVG task. Worth reading before you migrate.”
DevNews, Bluesky · 06:06 UTC
“The frontier AI model race has entered its comparison shopping phase.”
input, Bluesky · 21:27 UTC
“we threw several lifetimes of work at the wall, so some of it was bound to stick.”
Vincent Carchidi, Bluesky, skeptic · 22:24 UTC