Unverified50% confidenceBenchmarkExact time
MicroGPT achieved a throughput of tens of millions of tokens per second (10M tps) on an Apple M5 chip.
1
Sources
50%
Confidence
Medium-term (~90 days)
Relevance
8/30/2026
First Seen
Valid until: 11/28/2026
Sources
Related Entities
Related Claims
VerifiedThe memory bandwidth of Apple M-series chips can reach 200–800 GB/s (depending on the specific model)63% similarUnverifiedOn the RTX 5090, using D-Flash speculative decoding can achieve a speedup of more than 3x, while the improvement on Mac M4/M5 is between 1.5x and 1.8x.59% similarUnverified典型微控制器的RAM仅有数十KB，主频不超过数百MHz59% similarUnverifiedAMD's official benchmarks claim Halo is 40% faster than Apple in token generation for GLM Flash 30B, 12% faster for Qwen 3.5 122B, 7% faster for GPT-OSS 120B, and 3 to 7 times faster in image generation than the Apple M4 Pro.59% similarUnverifiedThe Apple M3 Max (128 GB unified memory) can run unquantized 70B models smoothly.59% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/823890API
curl https://kongchang.com/api/v1/knowledge/claims/823890MCP
get_claim(id=823890)