Taz

wiggling parameters

[Taz @taz_ca]

@tarat_211 @opencode how many tps are you getting?

[Taz @taz_ca]

@arcee_ai @ENERGY since it's voluntary data contributions, it would be awesome if u guys considered open sourcing it after

it could really push the field a bunch

[Taz @taz_ca]

@clarejtbirch great job on this small boi too!

do you guys recommend working with inkling small in NVFP4 too? or is the performance degradation more pronounced at the smaller size

[Taz @taz_ca]

@Presidentlin u tha real whisperer, i saw u mention deepseek a few hours ago lmao

[Taz @taz_ca]

@luke_chaj could be the worst outcome ever if that plays out

[Taz @taz_ca]

huge slap to open source models 😭

they gotta fight for relevance against this paper thin margin delivery while tryna compete on intelligence

[OpenAI @OpenAI]

We are committed to pushing the model frontier across cost efficiency, capability, and speed.

Starting today, we are reducing prices for GPT-5.6 Luna by 80% and GPT-5.6 Terra by 20% , and offering a faster option for GPT-5.6 Sol in the API.

Luna and Terra’s lower prices are reflected in how usage is counted in Codex and ChatGPT Work, so your usage goes further.

[Taz @taz_ca]

@sama now if u can make 5.7 luna hit 5.6 sol level intelligence, that's actually gonna be the promised intelligence too cheap to meter we all want~

[Taz @taz_ca]

@luke_chaj its a double edged sword because either way consumers will win with cheaper prices

[Taz @taz_ca]

@Shekswess @latkins @lokahq @arcee_ai wow it looks impressive, nice one!

why did u guys pick trinity mini in particular?

[Taz @taz_ca]

@jun_song they are very likely disaggregating the processing with amd chips vera rubin style, there's no 44gb limit

[Taz @taz_ca]

@EthanLipnik but codex’s autocompaction is so good u could forget context limits are a thing

[Taz @taz_ca]

@EthanLipnik soon soon, they’re just waiting for the hoarders to run out so not everyone smashes the server at the same time

[Taz @taz_ca]

@richychn @FireworksAI_HQ @Kimi_Moonshot @vercel @OpenRouter interesting but that would explain the latency but not the lower TPS right?

[Taz @taz_ca]

@ziv_ravid 50th like, start writing buddy 🥊

[Taz @taz_ca]

@thsottiaux do i smell a price cut or a weekly usage boost? 👀

[Taz @taz_ca]

@totheagi woohoo, can't wait to run this setup at my local nuclear reactor 🔥

[Taz @taz_ca]

@luke_chaj @apathium65906 @dhtikna 10T is rumours so can’t say but Mythos and fable might be there or even smaller actually given their leaked margins

they’ve got some insanely compressed intelligence on those boys for sure

[Taz @taz_ca]

@iamgrigorev mannn hype, the model factory idea is so neat

super excited to see what yall cook next 🫶

[Taz @taz_ca]

@luke_chaj @apathium65906 @dhtikna yup something like that ty, can confidently say it’s a huge factor 🙌