Unverified50% confidenceBenchmarkExact time
In an FP16-precision 4K×4K×4K transposed matrix multiplication task, the M5 Ultra with Neural Accelerator enabled surpassed 107 TFLOPS, which is 3.51 times its own non-accelerated performance and 4.79 times that of the M3 Ultra.
1
Sources
50%
Confidence
Medium-term (~90 days)
Relevance
9/23/2026
First Seen
Valid until: 12/22/2026
Sources
Related Entities
Related Claims
VerifiedThe M4 chip's Neural Engine delivers a peak compute performance of 38 TOPS (38 trillion operations per second)74% similarUnverifiedRTX 3060的INT4 Tensor Core理论峰值算力约为101 TOPS，而FP16仅约12.7 TFLOPS，差距达8倍以上73% similarUnverifiedThe M5 Ultra's peak AI compute performance is more than 4x higher than that of the M3 Ultra.71% similarUnverifiedVia --quantization quark_mxfp4, SGLang dequantizes NVFP4 weights and requantizes them to MXFP4 at load time, achieving GSM8K accuracy recovery of 97.5%–100.2% on models such as MiniMax-M2.7 and GLM-5.1.68% similarUnverifiedNVFP4量化方案相比BF16基线实现了1.5倍的推理加速，同时在top-1准确率上保留了92%到97%的水平67% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/945110API
curl https://kongchang.com/api/v1/knowledge/claims/945110MCP
get_claim(id=945110)