SambaNova (@SambaNovaAI)
Agentic AI가 기존 추론 트레이드오프를 깨며 에이전트당 속도와 대규모 처리량을 모두 달성할 수 있다고 주장합니다. 독립 벤치마크(ArtificialAnlys)에 따르면 SambaNova가 DeepSeek 같은 모델에서 두 가지 성능 특성을 모두 제공하는 것으로 확인되었다는 내용입니다.

Agentic AI breaks the old inference tradeoff. You shouldn’t have to choose between speed per agent and throughput at scale. Independent benchmarks from @ArtificialAnlys confirm SambaNova delivers both on models like DeepSeek. Read more 👇 https://t.co/l6LIRt6XEN
Intel is set to acquire AI‑chip pioneer SambaNova at a valuation below its 2021 $5 billion peak. The deal, backed by Intel Capital and SoftBank, could reshape open‑source AI hardware and boost collaborations with Stanford and Palo Alto innovators. Read more to see what this means for the community. #Intel #SambaNova #AIChip #IntelCapital
🔗 https://aidailypost.com/news/intel-moves-acquire-sambanova-valuation-below-2021-usd-5-billion-peak
8k is nothing, but let's keep trying.... 💪
#SambaNova Cloud Launches the Fastest #DeepSeek-R1 671B 🚀🚀🚀🚀🚀
🔹 Full DeepSeek-R1 671B model now running at record 198 tokens/second/prompt on #SambaNova's RDU hardware, verified by Artificial Analysis
#SambaNova Cloud's Developer Tier is now live! 🚀 🚀 🚀
offering token-based pricing with an initial $5 credit (over 30 million Llama 8B tokens) expiring in three months.
• Rate limits have increased, especially for 70B and 405B models (50% increase), and Llama 8B (doubled)
Компанія SambaNova Systems представила демо власної великої мовної моделі на Hugging Face, позиціонуючи себе як сильного конкурента моделі o1 від OpenAI.
#AI #Llama #LLM #SambaNova #ШІ
https://thetransmitted.com/ai/sambanova-kydaye-vyklyk-modeli-o1-vid-openai/
Компанія SambaNova Systems, виробник ШІ-чіпів, представила нову велику мовну модель (LLM) Samba-CoE v0.2.
https://thetransmitted.com/ai/sambanova-anonsuvala-samba-coe-v0-2-yaka-perevershuye-konkurentiv/