Business Ideas/Semiconductors & chips/India
A flagship phone tops out near 19 tokens a second on a small model, and memory sets that limit, not the AI chip. App teams planning on-device AI features do not know which of theirs will fit.
Research Radar/5G & 6G/India
A survey of tiny models for 6G reports one study that turned a 4.55 megabyte neural radio protocol into a 1 kilobyte rule set, a 99.98 percent cut with no loss in task performance.
Research Radar/Semiconductors & chips/India
A review of AI chips shows memory bandwidth, not arithmetic, sets the speed of language models. A flagship phone tops out near 19 tokens a second, and an H100 near 24 on a 70 billion parameter model.