Lite

📈Build in public, grow in public.
Product Lead at AI startup
Open for chat📩, excited to connect.

[PixVerse @PixVerse]

Meet PixVerse R2, our new real-time world model.
Explore living worlds. Control and edit them with prompts. Shape the story. Meet characters that remember and respond.

[Lite @xtralite_ai]

MiMo-V2.6-Pro and Grok 4.7 (xhigh) both score 46 on Artificial Analysis.

but MiMo runs at ~130 t/s and costs $0.13 per task. Grok runs at ~39 t/s and costs $3.74 per task, nearly 29× more.

MiMo just set a new standard again. it feels like the daily workhorse you wake up and actually use.

Grok 4.7 only gained 2 points over 4.6, while output speed dropped to 39 t/s. the huge fast-and-good edge Grok built in the 4.5 era is gone.

@elonmusk , keep up. i really liked the Grok 4.5 experience, but 4.5 just isn't capable enough anymore.

[Lite @xtralite_ai]

pay attention guys. reset doesn't refresh instantly inside your current cycle. it defers the natural reset day.

be careful it overlaps with, or gets eaten by, your original refresh window. use it too close to that day and the value basically goes to zero.

[Lite @xtralite_ai]

funniest thing this weekend. this is true AGI. why do you need to earn $80 when you have a banked reset from @thsottiaux

[FomskyWei @Fomsky_Wei]

好了睡前我想到了一个绝妙的点子
然后我觉得我貌似成全网第一乐子人了..我现在真的很迷茫
我觉得这个事件会和某mac电鸡事件一样陪伴我很久了

你想办法自己去赚80美元重置一下,谢谢

卧槽codex你tmd给我退钱.今晚又睡不着了.
去你大爷的退钱!!!我让你自己去赚钱啊草!!!
我现在很迷茫.谁能救救我.

[Lite @xtralite_ai]

everyday you wake up, there's a new update from @fal × @MiniMax_AI 🫠

[Lovis Odin @OdinLovis]

I did this with GPT-6 Astra and a modify version of H3 Max Director from @fal , still in prototype for this ( little chunk )

[Lite @xtralite_ai]

Gemini: hello? it's me

[Z.ai @Zai_org]

We’re still up.

[Lite @xtralite_ai]

seems everything is down now.

what's happening with all AI companies? is there any shared infra down right now?

[Lite @xtralite_ai]

it's happening with Codex as well. any infra down right now?

[Lite @xtralite_ai]

what's happening? My Grok, Grok Bot, Claude seem to be overloaded and down for a while, being really slow or broken at sometime, anyone else like me?

[Lite @xtralite_ai]

what's happening? My Grok, Grok Bot, Claude seem to be overloaded and down for a while, being really slow or broken at sometime, anyone else like me?

[Lite @xtralite_ai]

realtime factor 0.28 is already fast. 5s of video in 1.4s of compute.

to my view, the real milestone is 5s landing in 1s, including transfer and network, so people actually feel the change after one second. that would be huge for live generated video.

[Blendi @BlendiByl]

realtime factor is 0.28, you can generate a 5s video in 1.4s now 🫨

[Lite @xtralite_ai]

Gemini 3.8 Flash and Muse Spark 1.2 both dropped. on Artificial Analysis they're fast and good now, Grok 4.6's old edge on that combo is basically gone.

@elonmusk, ship Grok 4.7. my SuperGrok Heavy is ready.

[Lite @xtralite_ai]

@oran_ge not sure it actually got cheaper. been looking at a lot of community posts, and it doesn't seem much cheaper. plenty of people saying the usage limit still burns through insanely fast.

[Lite @xtralite_ai]

ngl, Claude Fable 5.1 just one-shotted my 5h limit, and I haven't seen the outcome yet. will have to wait another 5h for it to continue.

eager to see how Astra performs @thsottiaux, I'll decide by then whether to renew my Claude subscription. don't let us down.

[Lite @xtralite_ai]

went through the 100 htmls. impressive as a batch of studies, but to my view, design still isn't a solved problem.

it's so hard to ship good design with actual taste. all 100 are just ok, not production-ready in a way that wins actual users.

[Mia @MiaAI_lab]

I've asked Claude Fable 5.1 to create 100 HTML files

Rules were simple:
- look stunning
- zero repetitive designs
- go full creative mode

All 100 files created in one single prompt.

These are the BEST results I’ve had with this type of experiment, beating any other model. It’s really good on the frontend, and there are almost no broken files.

It's truly impressive.

See the results for yourself (incl. prompts):
https://t.co/I69ecRBfwK

There are a LOT that I like, had to split into two pages:

Jumpgate - Hyperspace Warp Tunnel
https://t.co/YLhvAscgTx

Playable Synthwave Synth & Visualizer
https://t.co/CHnEisvbA7

Luxury Watch
https://t.co/T3Jfwf0XEr

Ether - Fluid Smoke Canvas
https://t.co/YXcg3bfIni

Signal Loss — Generative Glitch Portrait
https://t.co/Hz5MDyGfvh

Lava Lamps
https://t.co/fbG5o3mRAY

Chaos Instrument
https://t.co/XNp7F0IFfW

Northern Lights over Lake Inari
https://t.co/Q7eYOKSyTI

Pixel Art Studio
https://t.co/4ztiHoeM3z

Spiral Galaxy
https://t.co/q89uqEIgUn

Honeycomb Hexagonal Navigation
https://t.co/wF95Tt9awM

More below 👇

[Lite @xtralite_ai]

Claude Fable 5.1 is out. all-time high on the intelligence index at 66. GPT-5.6 Sol at 61, Kimi K3 and GLM-5.3 at 60.

but cost went to $3.69 per task. that's ~3.9× GPT-5.6 Sol, ~4.4× Kimi K3, ~5.4× GLM-5.3.

@claudeai said cache reads cost 75% less, so it should be ~25% cheaper than Fable 5 (up to 45% on agentic work). AA still has Claude Fable 5.1 at $3.69 vs Fable 5 at $3.14. doesn't look cheaper from here.

curious what people are actually seeing on the bill so far.

[Claude @claudeai]

We’re introducing Claude Fable 5.1 and Claude Mythos 5.1.

They're the world’s most advanced models for coding and knowledge work.

[Lite @xtralite_ai]

@ErwanSoumhi @fal @levelsio next step is latency. visuals actually responding to what people type or say, not catching up a few seconds later.

after that the only limit is people's own imagination (and how fast they type or talk lol)

[Lite @xtralite_ai]

the h3 max wave on @fal is real. people are already building on it. https://t.co/ZjqadqGatv from @levelsio, https://t.co/lNhMnPNB4B from @fal.

just checked the stream. audio is pretty nice. it almost makes sense if you only listen (but don't actually watch)

video is definitely degraded to keep latency acceptable, 480p or less. faces are blurry, with that melted distortion.

that's the trade. you get a new format of content that generates itself, and lets people in while it's still generating. not a finished tape that pros made in advance.

cost isn't negligible if you actually run the numbers. $0.08/sec is ~$288/hour, ~$6,912/day to keep a 24h stream up.

curious what this becomes when that cost goes to nearly zero.

[Lite @xtralite_ai]

we'll have two separate paths for video / world gen, for a short while.

one is better quality, more authentic performance.
the other is how people actually interact with the content.

right now you're a one-way audience. things get made in advance, mostly by people who already know how to make great content.

later you'll get inside the process. you'll participate in how the content is produced. you can tweak it even after it already exists.

solaris sits on that second path. interfaces, real time, no code, you can move around inside what just got generated. still a long way from content production though.

these two will merge to one at some point. that's when the whole content industry gets rewritten.

[Runway @runwayml]

Today, we're sharing new research on Solaris, our first Interface World Model.

Solaris is a new kind of operating system that generates interactive interfaces frame by frame, in real time, with no code. We find that Solaris outperforms frontier LLMs when generating new interfaces, across structural similarity and information retention. Read more and request early access at the link below.

[Lite @xtralite_ai]

@pankajkumar_dev since when did they stop working on the pro models?

flash is indeed super fast. just wish 3.8 flash actually improves quality, not another speed bump.

[Lite @xtralite_ai]

early access, for free 🫡

[PixVerse @PixVerse]

No studio. No editor queue.
Paste a product URL into Shoplift by PixVerse and ship platform-native ad videos in minutes
—built for DTC teams running constant creative tests.
Free early access : https://t.co/ySTQ28ogCx
RT+Follow+Reply=300Creds(72H ONLY)