Radek Kotowski

LLM Inference @DeepInfra
Prev: Eng Lead @21shares, Head of Eng @Certara, Eng @imoim, waves are cool

[DeepInfra @DeepInfra]

Retrieval is where AI agents live or die. Miss the right context and the agent reasons from the wrong place.

@nvidia Nemotron 3 Embed is now on DeepInfra with Day-0 support, - the leading, embedding model for enterprise search. RAG, and agentic retrieval. Tops the RTEB leaderboard.

Two sizes, live now 👇

[Radek Kotowski @101radek]

@Jason @mati @theallinpod great interview Jison!

[DeepInfra @DeepInfra]

Big milestone for DeepInfra 📷

We've opened our first international #datacenter in Toronto!

The new facility will host 1,000+ NVIDIA Blackwell B300 GPUs, expanding our production-scale AI inference capacity while bringing lower-latency infrastructure closer to customers.

As #AI moves from training to production, inference infrastructure matters more than ever.
https://t.co/RN68zQe4fF

[DeepInfra @DeepInfra]

🆕Open models just hit frontier-level agent performance.

@LangChain tuned their Deep Agents harness for NVIDIA Nemotron 3 Ultra, top agent accuracy among open models, ~10x lower cost than closed alternatives.
It's live on DeepInfra now.

Point Deep Agents at your endpoint and go 👇

[Radek Kotowski @101radek]

our backrooms are filled with these

[DeepInfra @DeepInfra]

This looks really awesome. Many flops, much #NVIDIA B300s, deep infrastructure.

[Radek Kotowski @101radek]

what do you mean "ban open source"?

[Bill Gurley @bgurley]

Possible scenarios (ranked in order):

1/ Your country makes the best cheapest products. You export, and your citizens also buy the best.
2/ Your country doesn’t make the best/cheapest; but your consumers enjoy them (Brazil below).
3/ Your country doesn’t make the cheapest/best AND your country blocks access to your market. So you pay more for worst products, and you don’t export. Plus your local companies become less globally competitive as they are protected.

#3 is the worst. It’s what Europe did when America thrived. And it’s what many in Washington recommend today.

[DeepInfra @DeepInfra]

Text to video in ~2.4 seconds.

FastWan-QAD-FP8-1.3B is now on DeepInfra 🆕

5-second 480p clips from a prompt, landscape or portrait. 3-step inference, $0.0025/second.

[Ahmad @TheAhmadOsman]

The tides are shifting.

[DeepInfra @DeepInfra]

🆕🚀Batch API is live on DeepInfra.

Run large inference jobs → embeddings, evals, classification, summarization — asynchronously at 20% off real-time pricing.

OpenAI-compatible. If you've used OpenAI's Batch API, your existing code just works. Change base_url, done.

Results back within 24h 👇

[DeepInfra @DeepInfra]

@Zai_org GLM 5.2 = GPT-5.5 coding performance
at 1/6th the cost.
open weights.
now $1 / $4 / $0.18 per 1M tokens
best price on the market 🔥

[Mat Velloso @matvelloso]

All day using GLM 5.2. Didn't miss much. First open model that passes the bar as a daily driver. Things are not going to be the same.

Damn, now I want to buy some serious hardware.

[Radek Kotowski @101radek]

@btcjvs @Keir_Starmer yes by ensuring there is less brainrot, thus making people more effective

[Rihard Jarc @RihardJarc]

A few thoughts that come up from this suspension of Fable/Mythos:

1. Even if this decision is changed the move set a precedence. Investors investing in AI model providers will start to price for this risk. The idea of one company monopolizing and developing “digital god” and reaping all the economic benefits of it is dead. This opens the question of the ability of these top frontier AI labs to pay for the big compute commitments.

2. Enterpises will rush to orchestration platforms as they will want to have the option to switch between different AI model providers with ease not just because of effectivness of models but also because of the risk that the model or provider is banned by the government. Hyperscalers are in a great position to be that orchestration layer between different providers.

3. There is going to be an acceleration of AI models exclusive for specific vertical use-cases (models that dont have cyber or other capabilities so the chances of gov reach is lower), competiton here is going to expand to outside of the top 3 AI labs IMO.

[Citrini @citrini]

The risk of the government deciding that a model is too dangerous should only add to the reasons why open source models running on local hardware can be a reasonable alternative.

[JFPuget 🇫🇷🇺🇦🇨🇦🇬🇱 @JFPuget]

I bet https://t.co/ZtrF8ZKDHu valuation jumped suddenly.

Same for any non US player.

The world is learning from this US admin how depending on US technology is a liability. As French I fully get it. As an employee of a US company I regret it.

This is bad for me a non US national, and this is bad for the US company that employs me.

[Radek Kotowski @101radek]

IPO is over, we can escalate again

[Edward Z. Yang @ezyang]

In the same way you should always "look at the data", I think you should also always "look at the source code"

[Radek Kotowski @101radek]

@alexatallah @cclark @OpenRouter does it come with faucet included?

[Radek Kotowski @101radek]

while on the surface this looks like a healthy trend, some of the reasons behind this shift are concerning.

[Andrew Yeung @andruyeung]

Clubbing is dead and has been replaced by fitness & wellness.

Ppl used to party to socialize and date but now they do things like HYROX, bathhouses, and running raves.

The death of clubbing is something to be studied:
— US has lost 12% of its nightclubs in the last 24 months
— 25% of US adults didn’t drink at all last year
— Gen Z drinks 30% less than Millennials did at the same age

On the flip side:
— According to Strava, the number of running clubs recorded on the platform increased 3.5x in 2025
— 72% of Gen Z go to run clubs to meet new people
— Sauna and spa market: $11.8B → $22.4B by 2034

The post-alcohol economy is gonna be a massive category.