FuriosaAI
@furiosa.ai
Our mission is to make AI computing sustainable, enabling everyone on Earth access to powerful AI.
Looking ahead, while the industry will continue to evolve in ways no one can fully predict, our focus is clear: making inference seamless, energy-efficient, and production-ready at scale. The next phase of AI inference is here. More to come from Furiosa in 2026 🚀
December 30, 2025 at 6:50 PM
Looking ahead, while the industry will continue to evolve in ways no one can fully predict, our focus is clear: making inference seamless, energy-efficient, and production-ready at scale. The next phase of AI inference is here. More to come from Furiosa in 2026 🚀
This year’s progress includes:
→ Real-world validation, delivering up to 2.25× better performance per watt compared to GPUs
→ Major SDK advances across multi-chip scaling and tensor parallelism
→ Introduction of the NXT RNGD Server
→ A $125M Series C bridge to accelerate our next phase of execution
→ Real-world validation, delivering up to 2.25× better performance per watt compared to GPUs
→ Major SDK advances across multi-chip scaling and tensor parallelism
→ Introduction of the NXT RNGD Server
→ A $125M Series C bridge to accelerate our next phase of execution
December 30, 2025 at 6:50 PM
This year’s progress includes:
→ Real-world validation, delivering up to 2.25× better performance per watt compared to GPUs
→ Major SDK advances across multi-chip scaling and tensor parallelism
→ Introduction of the NXT RNGD Server
→ A $125M Series C bridge to accelerate our next phase of execution
→ Real-world validation, delivering up to 2.25× better performance per watt compared to GPUs
→ Major SDK advances across multi-chip scaling and tensor parallelism
→ Introduction of the NXT RNGD Server
→ A $125M Series C bridge to accelerate our next phase of execution
Over the past year, we’ve moved from architectural conviction to deploying inference systems built for what’s next. Working with partners like LG AI Research and OpenAI, we’ve validated a core belief: the future of AI acceleration starts with rethinking the fundamental compute primitive.
December 30, 2025 at 6:49 PM
Over the past year, we’ve moved from architectural conviction to deploying inference systems built for what’s next. Working with partners like LG AI Research and OpenAI, we’ve validated a core belief: the future of AI acceleration starts with rethinking the fundamental compute primitive.
Inference has quietly become one of the largest cost centers behind the adoption of agentic AI. As models scale and workloads move from experimentation to production, the economics of inference are forcing organizations to rethink infrastructure to keep AI practical, reliable, and sustainable.
December 30, 2025 at 6:49 PM
Inference has quietly become one of the largest cost centers behind the adoption of agentic AI. As models scale and workloads move from experimentation to production, the economics of inference are forcing organizations to rethink infrastructure to keep AI practical, reliable, and sustainable.
Come visit our poster and learn more about our commitment to deep research with real-world impact 🌎
Read the paper here: arxiv.org/abs/2507.06996
Read the paper here: arxiv.org/abs/2507.06996
December 3, 2025 at 8:13 PM
Come visit our poster and learn more about our commitment to deep research with real-world impact 🌎
Read the paper here: arxiv.org/abs/2507.06996
Read the paper here: arxiv.org/abs/2507.06996
We’re excited to carry this momentum into 2026 as RNGD moves toward broader availability.
www.crn.com/news/compone...
www.crn.com/news/compone...
December 1, 2025 at 7:47 PM
We’re excited to carry this momentum into 2026 as RNGD moves toward broader availability.
www.crn.com/news/compone...
www.crn.com/news/compone...
Learn more and register here: luma.com/hes7s3du?tk=...
November 7, 2025 at 10:37 PM
Learn more and register here: luma.com/hes7s3du?tk=...
To showcase this, we're powering a live multilingual demo, representing all 21 APEC member economies, on OpenAI's gpt-oss-120b model running efficiently on just two RNGD chips.
Interested in the future of efficient compute? Come talk to our team to learn more.
Interested in the future of efficient compute? Come talk to our team to learn more.
October 31, 2025 at 6:32 AM
To showcase this, we're powering a live multilingual demo, representing all 21 APEC member economies, on OpenAI's gpt-oss-120b model running efficiently on just two RNGD chips.
Interested in the future of efficient compute? Come talk to our team to learn more.
Interested in the future of efficient compute? Come talk to our team to learn more.
Read the story here 👉https://furiosa.ai/blog/serving-gpt-oss-120b-at-5-8-ms-tpot-with-two-rngd-cards-compiler-optimizations-in-practice
October 17, 2025 at 6:05 PM
Read the story here 👉https://furiosa.ai/blog/serving-gpt-oss-120b-at-5-8-ms-tpot-with-two-rngd-cards-compiler-optimizations-in-practice
Performance gains vs. 2025.2.0:
✅ Llama 3.1 8B: Up to 4.5% average throughput boost and up to 55% average reduction in Time-to-First-Token (TTFT)
✅ Llama 3.1 70B: Up to 3× average throughput improvement and up to 35% TTFT reduction
✅ Llama 3.1 8B: Up to 4.5% average throughput boost and up to 55% average reduction in Time-to-First-Token (TTFT)
✅ Llama 3.1 70B: Up to 3× average throughput improvement and up to 35% TTFT reduction
October 10, 2025 at 7:11 PM
Performance gains vs. 2025.2.0:
✅ Llama 3.1 8B: Up to 4.5% average throughput boost and up to 55% average reduction in Time-to-First-Token (TTFT)
✅ Llama 3.1 70B: Up to 3× average throughput improvement and up to 35% TTFT reduction
✅ Llama 3.1 8B: Up to 4.5% average throughput boost and up to 55% average reduction in Time-to-First-Token (TTFT)
✅ Llama 3.1 70B: Up to 3× average throughput improvement and up to 35% TTFT reduction
Read the full announcement here: www.bytebt.com/bytebridge-s...
October 7, 2025 at 6:59 AM
Read the full announcement here: www.bytebt.com/bytebridge-s...