@SpaceXAI: @grok multimodal, imagine | prev @GoogleDeepMind: gemini & imagen
Earth
- Tweets 1,027
- Following 563
- Followers 6,238
- Likes 3,642
Pinned Tweet
Thrilled to receive this feedback from Tesla — they tackle the world’s most critical multimodal challenges.
Extremely proud that the team’s work is paying off🚀
Grok 4.6 multimodal is a step change from Grok 4.5. It’s one of the under-discussed improvements, and I’ve been very impressed by it.
My daily work includes reviewing lots of videos and understanding the context; Grok 4.6 improves the productivity of such workloads by at least 10x if not 100x.
Such workflows may not be captured by common VLM benchmarks, but in my use cases it outperforms Gemini 3.5 and Gemma 4, which is considered the SOTA of VLMs in my opinion.
Hats off to the multimodal teams—you did a great job.
51
150
1,235
757,628
Really looking forward to seeing first gemini response from the space 🚀
17
4,070
Hexiang (Frank) Hu retweeted
We haven't forgotten about Grokipedia. v0.2 will be better than ever.
313
280
6,063
265,602
RT @karaebel: Very cool to work with @PJaccetturo and his team at Genre AI on this pilot scene of the Odyssey.
If you have any questions a…
2
44
Hexiang (Frank) Hu retweeted
xAI's Grok Imagine Image 2.0 takes #4 on the Artificial Analysis Text to Image Leaderboard, the highest ranked model outside OpenAI, a 14 place climb over the previous generation, and lands on the Pareto frontier for quality vs price
Grok Imagine Image 2.0 is xAI's latest image model released in August. The model supports text to image, image editing, as well as multi-reference generations with up to five input images.
Grok Imagine Image 2.0 is a large step up on the previous generation grok-imagine-image-quality: it climbs from #18 to #4 in Text to Image and from #16 to #10 in Image Editing, closing the gap to the leaders on every Text to Image capability and use case we measure, with the largest gains in Knowledge, Text Rendering and Lighting.
Grok Imagine Image 2.0 is available through the xAI API as grok-imagine-image-2.0, in Grok Imagine on grok.com and the Grok apps, and through fal and Replicate.
Congratulations to @SpaceXAI and @elonmusk on the release!
See below for our analysis and example outputs of Grok Imagine Image 2.0 in the Artificial Analysis Image Arena 🧵
24
29
424
43,330
Hexiang (Frank) Hu retweeted
building models that understand us requires building models of us - super excited to share some work in that direction and hope people find it useful!
For AI to work with us, it needs to understand us
Today, we're introducing Persimmon, the first large-scale model designed to realistically simulate how people talk and interact
12
20
240
19,766
Huge congrats on the new model launch @ericzelikman @YuchenHe07 🚀💫🌟
For AI to work with us, it needs to understand us
Today, we're introducing Persimmon, the first large-scale model designed to realistically simulate how people talk and interact
1
13
2,415
LFG
Image + video + smart agents work better together at grok.com/imagine/agent
A solid step in agentic video creation — we’d love your feedback!
10
2,032
Looped transformers aren’t a new species. They’re a way to buy FLOPs without buying parameters.
If intelligence tracks useful compute more than weight count, that’s the correct trade.
Stop blowing up Looped Transformers — it's not magical, it doesn't add recurrence to the (still and always) feedforward network, it was done years ago (Dehghani et al., 2019). All it does is making your model deeper, the same as if you added more layers to begin with.
1
16
2,375
Hexiang (Frank) Hu retweeted
This is why companies that don’t pull the plug on you matter: OpenAI will cut Cursor after the acquisition, then keep using X to promote their models.
The asymmetry is the whole story.
1
1
14
889
Hexiang (Frank) Hu retweeted
Providers start an API business before they know which apps will matter. Cursor was one of those apps, used the API well, and the provider made a lot of money off it.
Asking why they “allowed” Cursor from day one gets the sequence backward. First comes the API, then the apps, then the scale.
The scary part isn’t Cursor. It’s the precedent: downstream startups now have to compete with the market and with their own provider (OAI in this case) flipping on them.
1
1
15
1,196