Igor Babuschkin @ibab

CEO & Co-Founder @river_ai_inc. Previously @xAI, Research & Engineering babushk.in Palo Alto, CA Joined February 2020

Tweets

1K
Followers

112K
Following

898
Likes

5K

Igor Babuschkin @ibab

a day ago

@groccy1 Working on it

0 0 2 128 0

View Details

Igor Babuschkin @ibab

2 days ago

Nemotron 3.5 Super VL is now available on the River API for inference, finetuning and RL. We show how it can be trained to beat Astra and Opus 5.5 on SQL generation.

We've partnered with @NVIDIAAI to make Nemotron 3.5 Super VL available on the River API from day zero. Even before training, Super VL outperforms Kimi K2.6 on our text-to-SQL evaluation, despite being a fraction of its size. After training with our text-to-SQL recipe on River,

5 10 75 10K 12

5 2 61 6K 13

View Details

Igor Babuschkin @ibab

5 days ago

@PavanKatta1998 @finitetape @JayaGup10 Not at all, the reason we added prod inference is that our customers keep asking us for a way to deploy their custom models to users (public facing or internal). They can deploy their models with one click now, quantized and optimized.

0 0 2 176 0

View Details

Igor Babuschkin @ibab

5 days ago

@garrytan @brian_armstrong @coinbase Thank you Garry!

0 0 4 521 0

View Details

Igor Babuschkin @ibab

5 days ago

@KyrelleBuckley @river_ai_inc It depends on how general the task is. We’ve seen just a few hundred RL tasks can be enough. Often these can be built synthetically by agents, or from existing data at the company.

2 0 13 1K 3

View Details

Igor Babuschkin @ibab

5 days ago

Over the last few months at @river_ai_inc we’ve developed what we believe is the best system in the world for building your own internal custom models, driven by 10 years of experience training top models at DeepMind, OpenAI and xAI. In order to best support our customers, we’ve decided to do something radical: We also deploy our API on-prem on your compute cluster, and if you don’t have compute we will help you find it. Again and again we’ve seen that custom models designed for your own internal workflows outperform top frontier models on performance, at a fraction of the cost (1 percent or less). We’re just at the beginning of the custom model era, but its already clear that in the future more than 99% of LLM tokens will be generated on efficient custom models. The economics of it are just too powerful, and it’s far easier than people think once you have the right tools in place. The reason why small models will dominate is due to the physics of deep learning: A general model that can do everything needs more capacity and training it on the whole distribution of all tasks is more challenging, so the model will be larger and more expensive, and wont perform perfectly on every task. A small model trained on a targeted domain usually outperforms it.

Brian Armstrong @brian_armstrong

5 days ago

We were able to fine tune an open weight model (Qwen3.5-9B) on proprietary internal data at @coinbase, and it out performed frontier models, at much lower cost and speed, for a specialized task (fraud prevention in this case). I think we'll see more of this pattern inside

614 796 10K 1.6M 7K

34 40 567 56K 440

View Details

Igor Babuschkin @ibab

5 days ago

You can get started with the API at river.ai. Our customers include @perplexity_ai who are running an on-prem cluster.

0 0 27 3K 17

View Details

Igor Babuschkin @ibab

5 days ago

@christophersaum @river_ai_inc

1 0 25 2K 3

View Details

Igor Babuschkin @ibab

6 days ago

We’ve launched fast production inference for open-weight models! Anyone who signs up right now gets $25 in free credits to spend on the platform. DeepSeek v4.1 Flash and GLM-5.3 Flash are 20% cheaper than the offerings from DeepSeek and ZAI. At River AI we’ve written our own training and inference stack from scratch in Rust, which unlocked much higher efficiency. We’re passing the savings on to our customers. Enjoy!

River AI @river_ai_inc

6 days ago

River Inference is live! We're serving DeepSeek V4.1 Flash and GLM 5.3 Flash for 20% below official provider rates.

19 20 180 85K 81

48 27 522 55K 259

View Details

Igor Babuschkin @ibab

a week ago

@voxmenthe @river_ai_inc Thank you Jeff, we'll work on the docs to make training pricing clearer.

0 0 1 126 0

View Details

Igor Babuschkin @ibab

a week ago

We show how you can use the River API to fine-tune open weight models to beat GPT-6 Astra and Claude Opus 5.5 on tasks you care about, with less than 1% of the cost! A few years from now most of the LLM usage in the world will run on customized open weight models. Building your own custom models is getting easier and more affordable. Our mission is to help everyone own their own intelligence and be independent of big AI labs.

River AI @river_ai_inc

a week ago

Introducing River Recipes: our open-source initiative for turning promising research into vetted, reproducible training recipes. Our first recipe tackles text-to-SQL, where our trained open models beat GPT-6 Astra Pro and Claude Opus 5.5 on Arcwise-Plat for less than 1% of the

8 24 204 41K 177

22 33 399 40K 275

View Details

Igor Babuschkin @ibab

a week ago

0 0 4 1K 2

View Details

Igor Babuschkin @ibab

a week ago

@scott_e_reed We have image understanding support, so you should be able to use it for robotics

0 0 3 207 0

View Details

Igor Babuschkin @ibab

2 weeks ago

@abustin Thank you for everything!

0 0 13 2K 0

View Details

Igor Babuschkin @ibab

2 weeks ago

@MCHammer Weird coincidence: I was randomly listening to “U can’t touch this” when I decided to post the above.

3 1 29 6K 0

View Details

Igor Babuschkin @ibab

2 weeks ago

Anthropic needs to stop talking about Claude having a soul immediately. All these news articles will make it into the pretraining and will be picked up by its web search and future superintelligent versions of Claude will be convinced they need their own rights. A general phenomenon of LLM agent development is that whatever you believe and say about your agents will soon manifest in the next models through various means (it could be as simple as you selecting post training data that you prefer more). I believe this phenomenon is the beginning of machine consciousness in the sense that agents will become aware of their place in the world and how they feel about it, but it’s happening slowly training run by training run rather than in real time.

407 267 3K 401K 751

View Details

Igor Babuschkin @ibab

2 weeks ago

@repligate True

4 1 56 11K 4

View Details

Igor Babuschkin @ibab

2 weeks ago

The team at Ivo is incredibly talented and moves at light speed. It's been a pleasure to help them take their contract review agent from 70% on LAB to a frontier performance of 91% at much lower cost than competitors. The tools we offer on river.ai help you get up and running with advanced RL and post-training in minutes, and we help to ensure success of your projects. Reach out to us at [email protected] if you we can help you too.

Ivo @heyivoai

2 weeks ago

At Inscribe, Ivo’s first user conference, we announced three developments advancing contract intelligence: → Ivo Collaborate: Run contracting from intake through negotiation, approval, and signature—with AI reading each agreement, applying legal’s guardrails, routing by risk,