Loading video...

Video Failed to Load

Go Home

it's crazy what a 1.5B model can do these days! "VibeThinker-1.5B is a 1.5-billion parameter dense language model. With a total training cost of only $7,800 USD, it achieves reasoning performance comparable to larger models like GPT OSS-20B Medium." runs perfectly on device!

MaziyarPanahi's profile picture

Maziyar PANAHI

17,903 subscribers

202,328 views • 7 months ago •via X (Twitter)

LIVE

Anya Rossi• Live Now

Private livecam show

0 Comments

No comments available

Comments from the original post will appear here

Related Videos

gpt-oss-20b passes the vibe check ✅ no way this is only a 20B model, it’s beating models 2–3x its size
0:16

gpt-oss-20b passes the vibe check ✅ no way this is only a 20B model, it’s beating models 2–3x its size

Flavio Adamo

256,252 views • 11 months ago

Here’s Deepseek r1 1.5B thinking through a problem — it’s comparable to 4o and Claude 3.5 Sonnet in a number of domains like math. Except… it’s a 1.5B model… and can run on virtually any hardware. Truly a huge efficiency leap.
0:13

Here’s Deepseek r1 1.5B thinking through a problem — it’s comparable to 4o and Claude 3.5 Sonnet in a number of domains like math. Except… it’s a 1.5B model… and can run on virtually any hardware. Truly a huge efficiency leap.

Aaron Ng

564,677 views • 1 year ago

DeepSeek R1 distilled to Qwen 1.5B easily runs on my iPhone 16 with MLX swift. Here's the 4-bit model reasoning entirely on device at almost 60 toks/sec:
0:19

DeepSeek R1 distilled to Qwen 1.5B easily runs on my iPhone 16 with MLX swift. Here's the 4-bit model reasoning entirely on device at almost 60 toks/sec:

Awni Hannun

1,139,656 views • 1 year ago

This is amazing! Someone trained a 1.5B model (with GRPO) to solve mazes! model:
0:12

This is amazing! Someone trained a 1.5B model (with GRPO) to solve mazes! model:

Victor M

139,407 views • 1 year ago

Free Blender AI Motion Capture Plugin: Runs a 1-billion-parameter model on a GPU with 8GB VRAM, supporting real-time capture and video uploads — no cost required.
0:10

Free Blender AI Motion Capture Plugin: Runs a 1-billion-parameter model on a GPU with 8GB VRAM, supporting real-time capture and video uploads — no cost required.

CYANPUPPETS

16,064 views • 1 year ago

Advice for AI engineers 💡 A small Audio Language Model transcribes audio in real-time on your device, without internet connection or cloud costs. For example, LFM2-Audio-1.5B by ↓
0:14

Advice for AI engineers 💡 A small Audio Language Model transcribes audio in real-time on your device, without internet connection or cloud costs. For example, LFM2-Audio-1.5B by ↓

Pau Labarta Bajo

31,993 views • 7 months ago

Grok 5 is currently training at the Supercomputer Colossus 2. Grok 5 trains with 6 trillion parameter and a much larger 10 trillion parameter model.
0:10

Grok 5 is currently training at the Supercomputer Colossus 2. Grok 5 trains with 6 trillion parameter and a much larger 10 trillion parameter model.

Testlabor

27,518 views • 2 months ago

You don't need a more expensive model. You need better reasoning. Every frontier LLM got smarter with SERV: - smaller ones match frontier performance - frontier ones go past what they were built for The question isn't which model to choose - it's what engine it runs on.
0:11

You don't need a more expensive model. You need better reasoning. Every frontier LLM got smarter with SERV: - smaller ones match frontier performance - frontier ones go past what they were built for The question isn't which model to choose - it's what engine it runs on.

OpenServ

29,525 views • 1 month ago

GPT-5.5 by Reasoning Effort: I've asked it in Codex to create a physics-based visualisation of RL cycles for different sized models (70b, 1t, 10t), to demonstrate how the amount of RL you can do differs by model size. My assessment of each: - Low: weird slop - Medium: kinda cooked - High: sort of tried but ultimately incoherent - Extra High: elite - really nice idea and well executed Obviously this is just one shot, but worth trying different reasoning levels for the new models, medium seems to be pretty good for GPT-5.5 and it was really bad for many previous GPT models.
0:15

GPT-5.5 by Reasoning Effort: I've asked it in Codex to create a physics-based visualisation of RL cycles for different sized models (70b, 1t, 10t), to demonstrate how the amount of RL you can do differs by model size. My assessment of each: - Low: weird slop - Medium: kinda cooked - High: sort of tried but ultimately incoherent - Extra High: elite - really nice idea and well executed Obviously this is just one shot, but worth trying different reasoning levels for the new models, medium seems to be pretty good for GPT-5.5 and it was really bad for many previous GPT models.

Peter Gostev (SF: 22-26 June)

209,258 views • 2 months ago

We’re thrilled to launch GPT-OSS—the 120B-parameter open-source LLM—on Phala Cloud! 🚀 🔒 Run the fully OSS model inside a secure TEE for end-to-end confidential AI. Explore the build 👉
0:13

We’re thrilled to launch GPT-OSS—the 120B-parameter open-source LLM—on Phala Cloud! 🚀 🔒 Run the fully OSS model inside a secure TEE for end-to-end confidential AI. Explore the build 👉

Phala

12,176 views • 11 months ago

You can now vibe code a language model. From a single prompt, GPT‑5.6 built the entire training pipeline and trained a model from scratch on my iMessage history. Locally on my Mac. It now generates replies in my writing style.
0:16

You can now vibe code a language model. From a single prompt, GPT‑5.6 built the entire training pipeline and trained a model from scratch on my iMessage history. Locally on my Mac. It now generates replies in my writing style.

Pietro Schirano

500,539 views • 16 hours ago

With the SAS prompt, we can now use language models like Gemini to learn from a robot's history. This allows the model to analyze parameter effects ⚡, and suggest ways to improve - similar to a real-life table tennis coach. 💡
0:08

With the SAS prompt, we can now use language models like Gemini to learn from a robot's history. This allows the model to analyze parameter effects ⚡, and suggest ways to improve - similar to a real-life table tennis coach. 💡

Google DeepMind

21,239 views • 1 year ago

gpt-oss-20b is lightning fast on a NVIDIA RTX 5090. I get around 160-180 tokens per second.
0:11

gpt-oss-20b is lightning fast on a NVIDIA RTX 5090. I get around 160-180 tokens per second.

Lewis Menelaws

127,103 views • 11 months ago

We released physics-intern: a simple harness for science problems! It gets models like Gemini 3.1 Pro to go from 17.7 -> 31.4, thus beating GPT 5.5 Pro. The physics-intern harness can wrap any model and via dedicated subagent boost the performance of the vanilla reasoning models. While I think more and more of these harness capability gains will be absorbed into the models (like prompting tricks disappeared over time) there is a lot to be gained right now by building good scaffolds for those models and integrating tools well. Interestingly, the exception we found that GPT 5.5 Pro actually didn't benefit from the physics-intern harness! Read more about it here: PS: I think the Harness[Model] notation is kind of nice.
0:12

We released physics-intern: a simple harness for science problems! It gets models like Gemini 3.1 Pro to go from 17.7 -> 31.4, thus beating GPT 5.5 Pro. The physics-intern harness can wrap any model and via dedicated subagent boost the performance of the vanilla reasoning models. While I think more and more of these harness capability gains will be absorbed into the models (like prompting tricks disappeared over time) there is a lot to be gained right now by building good scaffolds for those models and integrating tools well. Interestingly, the exception we found that GPT 5.5 Pro actually didn't benefit from the physics-intern harness! Read more about it here: PS: I think the Harness[Model] notation is kind of nice.

Leandro von Werra

96,827 views • 1 month ago

Fin-R1 is out on Hugging Face A Large Language Model for Financial Reasoning through Reinforcement Learning
0:12

Fin-R1 is out on Hugging Face A Large Language Model for Financial Reasoning through Reinforcement Learning

AK

89,100 views • 1 year ago

Introducing Alpamayo 1.5. Based on community feedback, we’ve updated our 10B-parameter chain-of-thought reasoning VLA model to be a more interactive and steerable engine for autonomous vehicle development. Built on the Cosmos-Reason2 VLM backbone, this release adds support for navigation guidance and flexible camera configurations while providing new post-training scripts for model adaptation. 🤗 Learn more:
0:11

Introducing Alpamayo 1.5. Based on community feedback, we’ve updated our 10B-parameter chain-of-thought reasoning VLA model to be a more interactive and steerable engine for autonomous vehicle development. Built on the Cosmos-Reason2 VLM backbone, this release adds support for navigation guidance and flexible camera configurations while providing new post-training scripts for model adaptation. 🤗 Learn more:

NVIDIA DRIVE

50,769 views • 3 months ago

yeah this is a bubble. $1M investment, $159 a device, for what you can perfectly do with a pocket notebook.
0:16

yeah this is a bubble. $1M investment, $159 a device, for what you can perfectly do with a pocket notebook.

ℏεsam

24,895 views • 1 month ago

NVIDIA Nemotron 3 Nano Omni, a new multimodal reasoning model, is now live on Jetson AI Lab and unifies vision, audio, and language into a single reasoning loop. 🙌 Power your NemoClaws by running this model with Ollama, vLLM and other inference frameworks on NVIDIA Jetson hardware. Try it ➡️
0:13

NVIDIA Nemotron 3 Nano Omni, a new multimodal reasoning model, is now live on Jetson AI Lab and unifies vision, audio, and language into a single reasoning loop. 🙌 Power your NemoClaws by running this model with Ollama, vLLM and other inference frameworks on NVIDIA Jetson hardware. Try it ➡️

NVIDIA Robotics

16,031 views • 2 months ago