Measured on device (edge-compat, spark-x2.5-1.7b-int4): Galaxy S26 · LiteRT-LM 0.16.0 · GPU · decode 17.8 tok/s · prefill 459 tok/s · TTFT 520 ms · all 1733 ops delegated (2026-09-07); Mac Studio M4 Max · LiteRT-LM 0.17.0 · GPU · decode 103.6 tok/s · prefill 2492 tok/s · TTFT 112 ms (2026-09-07); Galaxy S26 · LiteRT-LM 0.16.0 · CPU · decode 13.6 tok/s · prefill 83 tok/s · TTFT 2.65 s (2026-09-07). Record: https://github.com/john-rocky/edge-compat/blob/main/cards/spark-x2.5-1.7b-int4/CARD.md Measured on device (edge-compat, spark-x2.5-1.7b-int8): Galaxy S26 · LiteRT-LM 0.16.0 · GPU · decode 15.8 tok/s · prefill 637 tok/s · TTFT 400 ms · all 1736 ops delegated (2026-09-07); Mac Studio M4 Max · LiteRT-LM 0.17.0 · GPU · decode 95.8 tok/s · prefill 2308 tok/s · TTFT 121 ms (2026-09-07); Galaxy S26 · LiteRT-LM 0.16.0 · CPU · decode 12.3 tok/s · prefill 165 tok/s · TTFT 1.37 s (2026-09-07). Record: https://github.com/john-rocky/edge-compat/blob/main/cards/spark-x2.5-1.7b-int8/CARD.md XHToken/Spark-X2.5-1.7B converted to the LiteRT-LM (.litertlm) format for on-device inference with Google’s LiteRT-LM runtime. Tested on litert-lm 0.17.0. Spark-X2.5 is the SparkLLM Team’s (iFLYTEK) open…
Модальности:
Генерация текста
Области применения:
Логика и рассуждение
Задача: Генерация текста
Автор: litert-community
Теги: litert-lm, litert, litertlm, on-device, edge, spark, reasoning
Лайков: 3 | Загрузок: 748
Описание основано на материалах HuggingFace. Перевод выполнен автоматически.
