Nvidia posts first Groq 3 LPX benchmark at 3,400 tok/s, spotlighting its $20 billion wager
AI Market Summary
Nvidia disclosed first third-party benchmarks for Groq 3 LPU-based LPX racks, showing ~3,400 tok/s on a 100k-context Gemma 4 31B test and framing a material inference-speed advantage versus alternatives. The results help validate Nvidia's large Groq acquisition and its heterogeneous GPU+LPU inference strategy, supporting confidence in datacenter AI platform differentiation despite questions around scaling to larger MoE models and competitor refresh cycles.
Impact level
● Medium
Affected assets
NCSKNVDA2USD/USDT-2.11%
AI Insight · NCSKNVDA2USD/USDTAI Insight
▲ Bullish
Trade now
⚠️ AI-generated insights are based on news content and are provided for informational purposes only. They do not constitute investment advice or represent the views of BingX. Investing involves risk. Please trade responsibly.
Nvidia has published its first benchmark results for Groq 3-based LPX racks, pointing to a sizable performance edge. Each Groq 3 LPU carries just 500 MB of memory, yet a 31B-parameter dense model can be distributed to fit within a single rack—roughly in line with the active-parameter counts seen in much larger models such as DeepSeek V3. The figures are being framed as early validation of Nvidia’s $20 billion acquisition bet on Groq’s LPU technology.
