Multiverse Computing Unveils Breakthrough: All CompactifAI Models Now Run on Intel Xeon 6 Processors

What we drew from this source

The claims Via News extracted from this document. We point to the source; we don't replace it.

• The CompactifAI-compressed model retained strong accuracy relative to the uncompressed baseline, with only minor variations observed on standard benchmarks.

  60% confidence

• At one concurrent user, the compressed model reduced processing time from 5,056.34 seconds to 2,598.22 seconds, a 48.6% latency reduction.

  60% confidence

• The CompactifAI-compressed Llama 3.3 70B model delivered an output throughput of 3.86 tokens/second and total token throughput of 7.81 tokens/second, improvements of 93.6% and 94.1% over the uncompressed baseline.

  60% confidence

• ITL, TPOT, and TTFT metrics showed substantial reductions: ITL mean fell 48.9%, TPOT mean fell 48.3%, and TTFT mean fell 46.6%.

  60% confidence

• At the highest concurrency level tested (256 concurrent users), throughput increased by 107.0% and latency decreased by 51.7%.

  60% confidence

Cited in these Via News reports

• Google Loses Two Top AI Researchers to Rival Startup as DeepMind Chief Steps Back →
• OpenAI Loses Another Top Executive as AI Industry Leans Into Trust-Building — With Its Own Numbers Still Unverified →