Inferact

@inferact

Building the future of inference through @vllm_project

following
followers

Inferact and Google reportedly make a preliminary TPUv7 Kimi K3 2.8T model submission

SemiAnalysis says Inferact led the submission end to end using the TorchTPU vLLM stack.

8 Sources, 11h ago, updated 11h ago

GB300 NVL72 wideEP optimization is claimed to deliver 7.2x better performance per TCO than every other AI chip

SemiAnalysis says the copper backplane expands the scale-up domain to 72, allowing each GPU to hold only some experts.

5 Sources, 6d ago, updated 6d ago

Inferact, Anyscale and Nscale plan a meetup space for PyTorch NA attendees

Inferact says attendees can grab coffee, get work done and connect with others interested in inference and vLLM.

2 Sources, 6d ago, first seen 6d ago

Nvidia Rubin vLLM inference claimed to have 3.2x better profit per gigawatt than GB300 NVL72

SemiAnalysis also claims up to 10x better performance per dollar than GB300 NVL72 on the vLLM engine.

7 Sources, 10d ago, updated 10d ago

Cohere and vLLM plan a Toronto meetup on open-source AI

Cohere says a vLLM lead maintainer will discuss the project's direction, while NVIDIA AI engineers will cover agentic debugging.

5 Sources, 18d ago, updated 18d ago

TPUv7 reportedly delivers 56% better performance than Nvidia GB200 NVL72 on Kimi K3

SemiAnalysis says vLLM maintainers showed TPUv7 can reach 700 tokens per second per user on Kimi K3 through megakernel optimization.

5 Sources, 26d ago, updated 26d ago

Kimi K3 is reportedly 2.2–2.8× faster on vLLM

Inferact says it co-led the optimization effort with Red Hat AI, Nvidia and Huawei.

4 Sources, 32d ago, first seen 32d ago

Google Cloud and Inferact partner to expand TPU support in vLLM

Inferact says the collaboration’s output will be open source, with a community program offering contributors shared TPU capacity and help from its core vLLM maintainers.

1 Source, 36d ago, first seen 36d ago

Featured in

Inferact and Google reportedly make a preliminary TPUv7 Kimi K3 2.8T model submission

GB300 NVL72 wideEP optimization is claimed to deliver 7.2x better performance per TCO than every other AI chip

Inferact, Anyscale and Nscale plan a meetup space for PyTorch NA attendees

Nvidia Rubin vLLM inference claimed to have 3.2x better profit per gigawatt than GB300 NVL72

Cohere and vLLM plan a Toronto meetup on open-source AI

TPUv7 reportedly delivers 56% better performance than Nvidia GB200 NVL72 on Kimi K3

Kimi K3 is reportedly 2.2–2.8× faster on vLLM

Google Cloud and Inferact partner to expand TPU support in vLLM

Featured in

Inferact and Google reportedly make a preliminary TPUv7 Kimi K3 2.8T model submission

GB300 NVL72 wideEP optimization is claimed to deliver 7.2x better performance per TCO than every other AI chip

Inferact, Anyscale and Nscale plan a meetup space for PyTorch NA attendees

Nvidia Rubin vLLM inference claimed to have 3.2x better profit per gigawatt than GB300 NVL72

Cohere and vLLM plan a Toronto meetup on open-source AI

TPUv7 reportedly delivers 56% better performance than Nvidia GB200 NVL72 on Kimi K3

Kimi K3 is reportedly 2.2–2.8× faster on vLLM

Google Cloud and Inferact partner to expand TPU support in vLLM