Be Novita AI’s first Developer Advocate, building our dev community through content, events, and code to democratize open-source AI. SF-based with competitive pay + benefits—apply here:
Novita AI
科技、信息和网络
San Francisco，California 348 位关注者
Deploy AI models effortlessly with our simple API. Build and scale on the most affordable, reliable GPU cloud.
关于我们
- 网站
-
https://novita.ai
Novita AI的外部链接
- 所属行业
- 科技、信息和网络
- 规模
- 11-50 人
- 总部
- San Francisco，California
- 类型
- 上市公司
地点
-
主要
获取路线
156 2nd
US，California，San Francisco
Novita AI员工
动态
-
Running WAN 2.1? As cloud GPUs provider, believe us—picking the wrong one's like putting whipped cream on chicken wings—disaster. From VRAM to compute power, here's your zero-nonsense guide to picking the right one ?? 1/ The 14B model needs serious muscle: With 8x GPUs, H100s crush it in 287.9s, while A100s take 389.7s. The twist? 1.3B version runs sweet on a single 4090. 2/ Want to know how RTX 4090, H20, A100 & H100 perform on different Wan 2.1 models? Detailed benchmarks + memory usage data ??https://shorturl.at/lROE6 4/ Buy or rent? A single H100 costs $30K ?? 4090 costs $2K to buy, but rent for $0.35/hr. H100? Only $2.89/hr. Start here??https://shorturl.at/x15Sj
-
-
Novita AI转发了
cu @ NVIDIA GTC afterparty with Novita AI @ Plug and Play Tech Center..
Headed to NVIDIA GTC this year? Join us for an intimate afterparty with Novita AI at Plug and Play Tech Center! We'll be hosting a panel and interactive discussion led by Zhiyu (Edward) Cheng from NVIDIA, Simon Mo from vLLM, and Chayenne Zhao from sgl-project. The best part – you get to hang out with Junyu and Pierce ?? Spots are limited! RSVP: https://lu.ma/gtc-afters
-
-
Novita AI转发了
Our powerful DeepSeek R1 Turbo is live on Hugging Face! ?? It's better in every way: 64K context + 16K max output + 30 throughput + 99.9% stability + ??20% cheaper! Try it out on Hugging Face's model page ?? https://lnkd.in/gGtVHRBH - ?? Shoutout to Julien Chaumond and the team to make it work!
-
Thrilled to be part of Langfuse (YC W23) ecosystem! It's the ???????? ?????????????? ???????? ???????????? ???????????? ????????????????. Your team can collaboratively debug & analyze LLM apps with Novita AI APIs—try Qwen QwQ 32B, DeepSeek R1 & V3 Turbo, and more. Iterate smarter ?? https://buff.ly/pcEdRi9
-
-
Our powerful DeepSeek R1 Turbo is live on Hugging Face! ?? It's better in every way: 64K context + 16K max output + 30 throughput + 99.9% stability + ??20% cheaper! Try it out on Hugging Face's model page ?? https://lnkd.in/gGtVHRBH - ?? Shoutout to Julien Chaumond and the team to make it work!
-
FYI, this event is free to attend, no GTC ticket required. Sign up now ?? https://lu.ma/gtc-afters
Headed to NVIDIA GTC this year? Join us for an intimate afterparty with Novita AI at Plug and Play Tech Center! We'll be hosting a panel and interactive discussion led by Zhiyu (Edward) Cheng from NVIDIA, Simon Mo from vLLM, and Chayenne Zhao from sgl-project. The best part – you get to hang out with Junyu and Pierce ?? Spots are limited! RSVP: https://lu.ma/gtc-afters
-
-
New small model QwQ-32B is here. It's game-changing! QwQ-32B, Qwen's new reasoning model that beats DeepSeek-R1 (671B) and OpenAI o1-mini in key benchmarks! ?? FYI no speed up for the video, it's really THAT fast! ?? $0.18/1M input & $0.2/1M output ?? 32K context window for complex reasoning tasks Try it yourself ??https://shorturl.at/33OQZ
-
Our new models ?? DeepSeek R1 & V3 TURBO are here - full 671B, but BETTER ??20% cheaper than before - R1-turbo: $0.7/2.5 per 1M tokens (input/output) - V3-turbo: $0.4/1.3 per 1M tokens (input/output) ??Power boost 64K context + 16K max output + 99.9% stability + 30 throughput Try it now ?? https://buff.ly/g88t08X
-