Wags

$CELL $CPUNK Whale

[Wags @Cellgoliath]

@cis_female @chipsandpol @turtleakg So 1000+tps and less than 500ms TTFT is terrible? Even on 1T + parameter models, You are just spewing shit as you speak, point real world examples, because evidence shows you as dead wrong

[Wags @Cellgoliath]

@Im_actuallyacat They haven't proved to beat them on anything yet. Maybe cost efficiency but not inference speed, not even close, when you compare models equally

[Cerebras @cerebras]

Gemma 4 31B is now available in Public Preview on Cerebras. Our first multimodal model runs at over 1,800 tokens/s for ultra-fast image and text workflows.
Give it a try: https://t.co/0APeoQtTPc

[Andrew Feldman @andrewdfeldman]

@cerebras is now running @GoogleDeepMind's Gemma 4 - the leading open-weight multimodal model - at 1,851 tokens per second in public preview.

This is 35x faster than a typical GPU endpoint.

Cerebras speed also translates into world class latency - Gemma 4 on Cerebras returns its first answer token inclusive of reasoning in 1.5 seconds, making Cerebras the only provider that lets Gemma 4 be used in real-time settings.

This is the power of wafer scale.

[Wags @Cellgoliath]

@chetaslua @Lentils80 Do not discuss it in public and boom blast it on X lol

[Wags @Cellgoliath]

@0xSero Best show i've watched in a long time, Love following the fan accounts and all the different theories

[zan @zanbuilds]

Wildlife rangers die each year from animal threats...

RANGA is Live Wildlife Ranger Incident Intelligence

A Gemma 4 + @cerebras multi-agent command center for poaching-risk triage.

It uses 7 agents:

Camera, Animal Detection, GPS, Weather, Poaching Risk, Alert, and Orchestrator.

Upload or capture a trail-camera image, add location context, and get a structured ranger incident report.

Live app: https://t.co/jwz0JqZ0Oc

Built for the Cerebras x Google Gemma 4 Hackathon.

@cerebras @GoogleDeepMind

[Wags @Cellgoliath]

@Sol_Survivr @cerebras Go on…?

[solidSF @solidSF]

we’ll be live streaming our 24hour entry in the cerebras hackathon

[Morgan @morganlinton]

Huge thanks to Cerebras for giving me early access to Gemma 4.

Kicking off a fun build with it, here's the idea I came up with, I'm calling it Veriform 🔽

RL-trained model that turns messy text into clean JSON, without all the wacky hallucination problems models can run into, oh and all running entirely locally on my four year old Mac with only 32GB of ram.

What I think makes this interesting (at least to me) is the grader. Gemma 4 on Cerebras judges every single attempt at ~1,500 tok/s, fast enough that I can afford a judge that checks each field against the source text instead of rubber-stamping anything that kinda looks like valid JSON.

Faithful judge vs lazy judge. Let's see how it goes.

The adventure begins, and yes, this is the kind of stuff I love doing on a Sunday morning 🤘

Oh and because a diagram probably does a better job explaining what I'm about to do, here ya' go.

[Wags @Cellgoliath]

@AIMoneyMaps You need to get up to speed, OpenAI verified Sol is running on cerebra’s at 750tps

[Andrew Feldman @andrewdfeldman]

Yesterday, @OpenAI announced GPT-5.6 Sol - their most powerful model ever.

It launches on @Cerebras in July. Frontier intelligence at unprecedented speed. Read about it here: https://t.co/LUQQAQbek6

[TheValueist @TheValueist]

https://t.co/rCD1CcP2qW

[Wags @Cellgoliath]

@AretardInvestor @CancelledJew @Bilalinvest They have the fastest inference chip by far and solved a hardware problem no one else has solved for 75 years. It's physically impossible for any other player to catch up with them on inference unless they are able to design there own wafer scale engine chip dancing around patent

[Wags @Cellgoliath]

@itsKarad @MParakhin They said 750tps will be in July

[Wags @Cellgoliath]

@bigswingingdong @chetaslua Well you’ve heard wrong

[Wags @Cellgoliath]

@cryptoanuran @aleabitoreddit You haven’t even seen the specs of the jalapeño chip, you still going to say that if they can run at only 200-300 tps? While cerebra’s is running 750+tps

[Wags @Cellgoliath]

@smartertrader Fastest inference bar none, no one will be able to compete with cerebra’s speed unless they create their own wafer scale engine equivalent (no one has ever done it successfully besides cerebra’s). Inference market to be the largest part of AI market. 0 chance?

[Wags @Cellgoliath]

@PrincipatusCap Until they can use it and see the results of it whilst keeping the high intelligence, I think people still think it’s to good to be true

[Daniel Steigman @trekedge]

Been in the GPT-5.6 Sol preview and what stands out is how proactively it delegates. The new ultra mode spins up subagents and picks up tasks I didn’t remember I needed. Really impressive work by the research team.