> GPT 5.6 sol announced > 2x lower cost > will become generally available soon
It's OpenAI's best model and is 2x lower cost!
This is Fable level intelligence at 25% of Fable's cost 💕
OpenAI just dropped a three-model family under the GPT-5.6 banner—Sol as the high-end agentic workhorse, Terra for balanced production loads, and Luna for fast cheap inference—each carrying a 90 percent markdown on cached input tokens while output rates stay put. The preview is gated to roughly twenty vetted partners for now, with wider API access promised in the coming weeks.
> GPT 5.6 sol announced > 2x lower cost > will become generally available soon
It's OpenAI's best model and is 2x lower cost!
This is Fable level intelligence at 25% of Fable's cost 💕
Sol, Terra, and Luna list input at five, two-fifty, and one dollar per million tokens, then slash the cached-input rate to one-tenth of those figures. The move targets repeated prompt prefixes without touching output pricing, which remains unchanged from prior flagships.
Public commentary positions Sol as matching Fable-level results at roughly a quarter of the cost, yet no independent head-to-head numbers or detailed benchmark tables have surfaced from the limited preview itself.
Positive users praised GPT-5.6's lower costs for making advanced AI more affordable and accessible to builders, while negative users viewed the intelligence claims as unverified hype and marketing.
No Digg Deeper questions have been answered for this story yet.
Correction - the top model is still $30 per 1M token
But they have a Sonnet-like version that is 2x lower
Still about 50% of Fable!
> GPT 5.6 sol announced > 2x lower cost > will become generally available soon
It's OpenAI's best model and is 2x lower cost!
This is Fable level intelligence at 25% of Fable's cost 💕
can someone explain why the differential pricing for input and cached tokens?

@bindureddy did they define "soon"?

@bindureddy We all know that model will be nowhere near fable model intelligence, common.

@bindureddy No, Sol is not 2x lower than 5.5

@bindureddy Would love to see how more expensive this is in comparison to frontier OS models.

@tinkerersanky Please check my thread

@deliprao Probably this

@bindureddy Hype price until we get access to it. Until then just marketing.

@bindureddy How its 1/4th? Its actually half of Mythos!

@bindureddy where does it say it will become generally available??

@bindureddy I think I saw it uses more tokens so not exactly 2x cheaper.

@bindureddy the cost curve on frontier models is moving faster than most roadmaps anticipated. Sol pricing is going to force some uncomfortable conversations at competing labs

@bindureddy real game changer is not just
intelligence
it’s making that intelligence affordable
and accessible

@bindureddy I wouldn't bet on "soon." OpenAI tied broader access to finishing a deal with the government, not to a date. Mythos 5 and Fable 5 have been stuck in that exact spot for two weeks now, no restoration yet.

@bindureddy "Generally available soon" that's a long shot, when they said "weeks". Looking forward to Chinese releases in the meantime.

@bindureddy OpenAI really said price war is on😂 Fable level smarts at 25% the cost is actually wild.

@bindureddy We will see

@bindureddy 5.6 Terra won't be Sonnet-like; it will be more powerful than Opus 4.8, about 50% of Fable.
Luna will be Opus-like and 5x cheaper

@bindureddy Great improvement