pomterre

swe | hacking something

[pomterre @pomterree]

If you bought a 20x Pro or any OpenAI plans for that matter, you're getting royally scammed. They say Anthropic is the devil, what the fuck is this then?

[pomterre @pomterree]

@itsrickyszn @claudeai Cherrypicking a random bench that has an OpenAI cohort around it is wild.

[pomterre @pomterree]

@ChrissGPT Adaptative reasoning + Opus 4.8 fallback; that edge doesn't give it justice. Fable is in ANOTHER league.

[pomterre @pomterree]

@techfrenAJ 🤣

[pomterre @pomterree]

@pondorasti @shawmakesmagic Please update 5.6 mapping; defaults to GPT 5 for some reasons. There should technically be no model
mapping hardcoding, just let it loose deterministically.

[pomterre @pomterree]

People forget Mythos is 4 month old AND 5.6 Sol still can't match Fable easily. The problem is Anthropic were heads down deep on the next frontier like a hibernating grizzly bear ready to pounce back.

[pomterre @pomterree]

@v10se @QCXINT_ i like mistral here more its sucking taxpayer money the most

[pomterre @pomterree]

@orienwu @kunchenguid I mean if you don't see that, I can't help you. ANT revenue is skyrocketing while you have OAI stresstesting bankruptcy + 5.6 did not justify it, dud of a model.

[pomterre @pomterree]

@Aero96193997 @kunchenguid Even if that was true, that's a 2 month gap. Fable > 5.6 Sol; it's not even closed btw, and 1/2 of Fable price, not 1/3. But Sol apparently takes in more turns to achieve similarly to Fable.

[pomterre @pomterree]

@scaling01 These monickers are just a facade to replace mini and nano model naming. 5.5 beats Terra and Luna fwiw.

[pomterre @pomterree]

GPT 5.6 Sol: $10/$60
Fable: $10/$60

One needs many more turns in than the other to achieve similarly.

Let that sink in.

[Theo - t3.gg @theo]

Apparently I'm allowed to talk about GPT-5.6 now?

It's a damn good model. Not quite as "smart" as Fable, but it is incredibly capable. Fixed all the problems I had with GPT-5.5.

It is incredibly determined. Will run for a day without even using a /goal. It understands subagents incredibly well and is great at orchestrating. It's super pleasant in use cases like OpenClaw and Hermes Agent. It knows iOS dev incredibly well.

It has rough edges too, but FAR fewer than 5.5 did.

For many things, gpt-5.6-sol will become my obvious default.

I will share a lot more soon 🫡

[pomterre @pomterree]

@davis7 Isn't that what makes a model better; to understand intent better with minimal turns. So, I don't get why we are downplaying the dud of a model that 5.6 is, in which it needs more turns to achieve similarly to Fable one turn.

[pomterre @pomterree]

At only the price tag of $94,000; you can host GLM 5.2 forever and "FREE". Get rid of those useless $200 Max subs, completely worthless and under the marginal propensity curve!

[Tulip King 🌷 @tulipking]

local 1T parameter models are now only $94,000. I don't think we've ever seen a cost curve get pushed to zero as quickly as AI. We live in a prosperity era

[pomterre @pomterree]

@v10se @LukeberryPi fable won gg

[pomterre @pomterree]

@Teknium Yea, let's completely ignore the OpenClaw mention too.

[pomterre @pomterree]

@ben_sage No one is 100% altruistic. If you think that, you're living out of reality.

[pomterre @pomterree]

It pays to track down insiders. Who woulda thought.

[OpenAI @OpenAI]

GPT-5.6 Sol, along with Terra and Luna, will launch publicly this Thursday.

We’re expanding preview access globally now.