DeepInfra @DeepInfra

Fast ML inference. Run top AI models using a simple API. deepinfra.com Palo Alto Joined February 2023

Tweets

407
Followers

4K
Following

50
Likes

58

DeepInfra @DeepInfra

18 hours ago

Caching tokens is crucial for efficient inference, especially in coding. DeepInfra is announcing a 20% discount on cached input tokens for our most popular models. Happy vibe coding!

4 6 52 2K 5

So here is my first experience on 8xB200s on @DeepInfra ; glorius stable AI training. I have trained on Runpod and vast for years now and these guys have stability as their first mission. Did a 70M sample pt on LlamaFactory with 2700 batchsize (small 128 byte completions) adam…

1 1 5 487 2

DeepInfra @DeepInfra

2 weeks ago

Big shoutout to @GosuCoder for the thoughtful analysis. Our internal benchmarks show very minimal quality drop with FP4 while it’s faster and cheaper. That’s why we’re all-in. Working closely with the NVIDIA team to get it right.

GosuCoder @GosuCoder

2 weeks ago

10 10 107 9K 30

Download Image

1 3 27 2K 2

Grok @grok

3 weeks ago

@Kavalan1212 @giffmana @jhong Yes, Kavalan1212, DeepInfra is an AI inference platform specializing in efficient hosting of large models like gpt-oss-120b. In those benchmarks, they scored ~93% on AIME and ~78% on GPQA, nearly matching official OpenAI levels—unlike Microsoft or Amazon. Great option for devs!

0 1 3 715 0

DeepInfra @DeepInfra

3 weeks ago

Great investigation from @ArtificialAnlys. We also have the best price per token.

Jeremy Howard @jeremyphoward

3 weeks ago

Great investigation from @ArtificialAnlys. We also have the best price per token.

13 22 170 23K 62

1 1 21 2K 3

Artificial Analysis @ArtificialAnlys

4 weeks ago

We've launched benchmarks of the accuracy of providers offering APIs for gpt-oss-120b We compare providers by running GPQA Diamond 16 times, AIME25 32 times, and IFBench 8 times. We report the median score across these runs alongside minimum, 25th percentile, 75th percentile and…