@Sentdex
Account based inUnited States
About this account
- Account based in
- United States
- Connected via
- United States Android App
Account-level information from X, not a live location or the device used for a specific post.
gpus and tractors. Neural networks from Scratch book: https://nitter.cf/t.co/hyMkWyUP7R https://nitter.cf/t.co/8WGZRkUGsn
Alabama, USA
Joined May 2013
- Tweets18.9K
- Following394
- Followers112K
- Likes23.3K
Pinned Tweet
No training, no sim2real gaps, just a prompt and a dream.
The problem in robotics today is every task is a complex combination of locomotion, manipulation, vision, planning, and a bunch of other things that fall apart once you diverge from training data or characteristics from sim. This is why you see so many demos with perfect lighting conditions and always the same, or precisely placed objects, specifically chosen for the task and demo. You never see these things deployed in the real world despite seeing demos that look convincing, because everything falls flat in the real world with real unseen conditions.
Every single deployment is a unique edgecase.
Here, it's just a multimodal LLM (DSV4.1F) doing a task it was never deliberately trained for, using cameras it's never seen, servos/motors it didn't learn on, in an environment it's never dealt with and a form factor never tested.
It is currently slow, this video is 10x speed, but I am very confident it can be sped up as is, but also this is basically the slowest this level of intelligence will ever be again.
The more I work with general purpose multimodal LLMs, the more I am realizing that ACT, VLAs, and all this other stuff is probably just a sideshow and a distraction, and it's just going to largely get eaten up by general purpose models that just happen to also do robotics.
I don't think it'll always be LLMs, but the fact that LLMs can do some of the most challenging robotics tasks tells me it's kind of over for a large class of robotics training.
This does not include RL, however, for low level things, like gait, reaching, and other capabilities or "skills" for hardware. RL is still a king here and you really do want these models to be super fast and super small. RL is very good here and safe I'd say.
Decision models like Clef, Jev...etc remain mostly useless imho for robotics. I look forward to someone proving out a useful recipe that works these things into robotics, but I don't currently see it.
Every example of local models going truly fast with high quality on rtx pro 6ks, spark/gb10s, or dgx stations either comes directly from this group or it's someone copying from this group.
I've learned a lot, and spent way too much money on hardware, because of these people.
Ready to supercharge your local AI hardware? 👀
Build an optimized docker for your @NVIDIAAI RTX 6000 Pro or DGX Spark at docker.local-inference-lab.a…
wait, there's a config for $20K that runs GLM 5.3 at 70 t/s? send details!
have you even said "thank you" once to claude???
Exclusive: Anthropic is updating its usage policy for the first time in over a year. The new rules prohibit sustained "abusive or cruel behavior" towards Claude & add new restrictions about propaganda campaigns, surveillance & weapon development. theverge.com/ai-artificial-i…
Robot used here is @sourccey from Vulcan Robotics. If youre interested in an actually useful robot to R&D on like this, definitely hit them up.
For a deeper dive into how all this is working at the moment: youtube.com/watch?v=dGkJe3aI…
Breaking: data mining company mines data
“.. Each hour, Muse updates its dossiers on you and the people you’ve mentioned in chats, messages, and emails that Muse has read. .. Even people who don’t use Muse are subject to this mapping process by other people’s Muse agents.”
@time $META
time.com/article/2026/10/06/…
France might still be losing to China.
But USA is losing to France.
Mistral has released Mistral Large 4, scoring 38 on the Artificial Analysis Intelligence Index; France is back to having the most intelligent model from outside the US and China
@MistralAI has released Mistral Large 4 in Research Public Preview, with plans to release the weights of the 1T parameter (49B active) model at the end of October. It achieves 38 on the Artificial Analysis Intelligence Index, comparable to GPT-6 Luna (max, 38) and DeepSeek V4.1 Flash (max, 39). It also achieves 50% on the Artificial Analysis Cyber Index, level with GLM-5.3-Flash and ahead of models such as Kimi K3 and DeepSeek V4.1 Flash (max).
Key benchmarking results for Mistral Large 4 Preview:
➤ Most intelligent model from outside the US and China: Mistral Large 4 Preview scores 38 on the Intelligence Index, comparable to DeepSeek V4.1 Flash (max, 39) and GPT-6 Luna (max, 38). This makes it the most intelligent model from outside the US and China, ahead of countries such as South Korea and the United Arab Emirates
➤ Level with GLM-5.3-Flash on cyber defense capability: Mistral Large 4 Preview scores 50 on the Artificial Analysis Cyber Index, level with GLM-5.3-Flash (50) and behind MiMo-V2.6-Pro (56). Once its weights are released, it will rank among the top three open weights models on the Cyber Index. Its strongest result is on CyberGym-E2E-AA, where it scores 82%, ahead of MiMo-V2.6-Pro (79%) and GPT-6 Luna (max, 78%)
➤ Over 4x the Cost per Task of similar-intelligence open weights models: Mistral Large 4 Preview costs $1.13 per Intelligence Index task with standard pricing of $1.36/$4.18 per 1M input/output tokens, with $0.14 per 1M cached input tokens. For the first two weeks, Mistral Large 4 Preview will be served at a 50% launch discount, bringing its Cost per Task down to $0.57. This is still more costly than GLM-5.3-Flash ($0.25) and DeepSeek V4.1 Flash (max, $0.27)
➤ Strong document and image reasoning: Mistral Large 4 Preview scores 19% on GDP.pdf, on par with MiMo-V2.6-Pro (19%) and behind Kimi K3 (22%). This is an 18-point improvement from Mistral Large 3, partly driven by improvements in their API, which now accepts 100 images per request, up from 8 for previous Mistral models
Key model details:
➤ Context Window: 512k tokens
➤ Multimodality: Text and image input, with text output
➤ Pricing: $1.36/$4.18 per 1M input/output tokens ($0.14 per 1M cached input tokens), with 50% off for the first two weeks ($0.68/$2.09)
➤ Availability: Research Public Preview on Mistral's API, with open weights planned for the end of October
So far, easily the coolest thing with the DGX station is adding a single RTX Pro 6K gives you a 60,000 tok/sec prefill with DeepSeek V4.1 Flash. That TTFT speed is intoxicating!
Using recipe found here:
github.com/original-el8/dgx-…
I cannot escape the connection of the Reflection 70B fiasco with Reflection the company.
Every time I see this name I get irked a little bit. Maybe it's just me but I wonder how many people remember the totally unrelated Reflection 70B model and associate those feelings consciously or subconsciously and it hurts this co's rep.
How much of that negativity exists in LLMs too I wonder?
I've been on the hunt for any decision language model that actually performs generally useful tasks with robotics, including vision based ones like clef.
repeatedly, the finding I have is that these models are not overall intelligent enough to replace specialized RL/ACT/VLAs, or larger general purpose multimodal LLMs, for the actual reasoning and understanding parts. Because I'm still needing to make a call to this style model, there really aren't any significant gains.
I have also tested converting GLM 5.3 Flash (a general purpose and decently large/intelligent multimodal LLM) to just softmax over logits (copying the SemIF recipe that does work well on smaller models and competes with jev...etc), and this *also* does not work well in robotics, because the GLM 5.3 Flash model, and others like it, are trained deliberately to get a lot of performance and intelligence (read: usefulness, effectiveness) from doing test time compute.
It has left me really wondering where this "decision" style of model will actually fit in.
I thought vision enabled variants could be possibly epic, but I just keep finding when I put them into real robotics pipelines, they just aren't better than other techniques.
Maybe we just still have yet to figure out how to best utilize these models.
If you've seen any cool (non-priv data) robotics example using these decision models, please share!
Deepseek V4.1 Flash is seeming like a pretty good robotics control model.
Happened to be testing DSV4.1F on DGX Station and it is so far much better than GLM 5.3 Flash at reading imagery and making robotics decisions, at least for the pick-place task I was trying.
economies always form. the items don't even need to be real. people have been trading in videogame items and gold for decades
In my opinion, the best model on the planet at the moment is GLM 5.3 Flash.
It's super fast, super smart for the size, and is within reach to run locally for many. Love working with GLM 5.3 Flash.
