#CentML - Twitter Hashtag

about 1 year ago

🚨 JUST IN 🚨 🇨🇦 $NVDA acquires Toronto-based AI startup CentML! 🤖💡 NVIDIA continues to strengthen its AI leadership with cutting-edge tech from the Canadian startup. #AI #NVIDIA #CentML #TechNews #AIRevolution #CryptoNews

CryptoNewsHntrs's tweet photo. 🚨 JUST IN 🚨
🇨🇦 $NVDA acquires Toronto-based AI startup CentML! 🤖💡

NVIDIA continues to strengthen its AI leadership with cutting-edge tech from the Canadian startup.

#AI #NVIDIA #CentML #TechNews #AIRevolution #CryptoNews https://t.co/Pgj8STC69N

0

7

2

0

161

Gennady Pekhimenko

@Dr_GGP

over 1 year ago

Amazing new collection of powerful @Meta #Llama4 models in partnership with @AIatMeta supported right away on CentML's Platform: https://t.co/asaa3jx2a4. #Llama4 #meta #centml

0

3

0

165

Gennady Pekhimenko

@Dr_GGP

over 1 year ago

Great job by the @CentML_Inc team with impressive results on the most popular @deepseek_ai R1 model. Optimization is the key! #DeepSeekR1 #inference #optimization #centml

Artificial Analysis

@ArtificialAnlys

over 1 year ago

Comparing DeepSeek R1 APIs: We are now tracking DeepSeek R1 APIs from @DeepInfra, @FireworksAI_HQ , @togethercompute, @nebiusai, @hyperbolic_labs, @novita__ai, @CentML_Inc and Microsoft @Azure As of today, we’re seeing the fastest speeds on: @togethercompute, @CentML_Inc and @Fireworks We’re tracking the best prices on: @DeepInfra and @nebiusai Microsoft Azure launched an endpoint but we note that it is in a free preview phase with no disclosed pricing, and currently appears to be limited to a 4000 token context window. DeepSeek’s first party API offering has been unreliable over the past two weeks, with the majority of our benchmarking workloads generating errors on most days. DeepSeek are also currently blocking API account balance top-ups. DeepSeek R1 is a particularly hard model to host compared to most other open weights models because it’s so large - at 671B total parameters, it cannot fit on a single 8xH100 node. It can be very expensive for providers to offer APIs at small scale. With so much time spent generating reasoning tokens, faster output speeds are highly material to user experience.

ArtificialAnlys's tweet photo. Comparing DeepSeek R1 APIs: We are now tracking DeepSeek R1 APIs from @DeepInfra, @FireworksAI_HQ , @togethercompute, @nebiusai, @hyperbolic_labs, @novita__ai, @CentML_Inc and Microsoft @Azure

As of today, we’re seeing the fastest speeds on: @togethercompute, @CentML_Inc and @Fireworks

We’re tracking the best prices on: @DeepInfra and @nebiusai

Microsoft Azure launched an endpoint but we note that it is in a free preview phase with no disclosed pricing, and currently appears to be limited to a 4000 token context window.

DeepSeek’s first party API offering has been unreliable over the past two weeks, with the majority of our benchmarking workloads generating errors on most days. DeepSeek are also currently blocking API account balance top-ups.

DeepSeek R1 is a particularly hard model to host compared to most other open weights models because it’s so large - at 671B total parameters, it cannot fit on a single 8xH100 node. It can be very expensive for providers to offer APIs at small scale. With so much time spent generating reasoning tokens, faster output speeds are highly material to user experience.

7

165

23

68

31K

0

8

1

0

357

Gennady Pekhimenko

@Dr_GGP

over 1 year ago

Today with the @CentML_Inc platform… 1. Name your endpoint 2. Choose your model 3. Select a hardware configuration 4. Deploy Launching an LLM has never been easier…even from your phone 😀 #centml #optimization #NVIDIA #gpu #google #TPU #AWS