eIQ Gen-AI Flow is a modular, end-to-end software pipeline optimized to bring large and small language models (LLMs and SLMs) to NXP applications processors like the i.MX 95, i.MX 93 and i.MX 8M Plus. The pipeline includes wake word, speech-to-text, text-to-speech and Gen-AI models to run on device CPU and/or NPU. Gen-AI Flow also includes retrieval augmented generation (RAG) to securely and easily fine-tune LLMs with custom, private knowledge sources like PDFs.
| Performance Tier | Platform | Configuration | TTFA Avg (s) | CPU Avg (%) | Memory Avg (MB) | STT WER* (%) | LLM TTFT (s) | LLM TPS (tok/s) | TTS RTF |
|---|---|---|---|---|---|---|---|---|---|
| High Performance+ | NXP FRDM-IMX95 15x15 + ARA240 | whisper-small.en + rag + Qwen2.5-Coder-1.5B (ARA240 DNPU) + tts | 3.47 | 64.49 | 2943.67 | 5.82 | 0.54 | 18.69 | 0.16 |
| High Performance | NXP FRDM-IMX95 15x15 | whisper-small.en + rag + danube-500M-q4 (with NPU) + tts | 3.77 | 47.95 | 5075.58 | 5.29 | 0.27 | 10.24 | 0.33 |
| moonshine-base + rag + tts | 1.23 | 33.95 | 1558.37 | 6.35 | N/A | N/A | 0.21 | ||
| High Performance | NXP i.MX952 EVK | moonshine-base + rag + danube-500M-q4 + tts | 3.76 | 66.11 | 2198.38 | 5.82 | 0.88 | 8.33 | 0.48 |
| Lightweight | NXP FRDM-IMX93 | moonshine-base + rag + tts | 2.24 | 54.62 | 1152.09 | 6.88 | N/A | N/A | 0.39 |
| moonshine-tiny + rag + tts | 0.83 | 52.61 | 1454.54 | 14.81 | N/A | N/A | 0.39 | ||
| Lightweight | NXP FRDM-IMX8MPLUS | moonshine-base + rag + Qwen2.5-Coder-1.5B (ARA) + tts | 2.84 | 70.36 | 2738.27 | 7.41 | 0.54 | 18.32 | 0.26 |
| Minimal | NXP i.MX8MM EVK | moonshine-base + rag + tts | 2.22 | 41.28 | 1317.72 | 6.88 | N/A | N/A | 0.54 |
| moonshine-tiny + rag + tts | 1.19 | 40.01 | 1601.84 | 14.81 | N/A | N/A | 0.58 | ||
| Ultra-Light | NXP i.MX91 11X11 EVK | moonshine-tiny + rag + text (no TTS) | N/A | 65.98 | 1039.03 | 8.47 | N/A | N/A | N/A |
3 ダウンロード
注: より快適にご利用いただくために、ソフトウェアのダウンロードはデスクトップで行うことを推奨します。
5 ハードウェア提供
追加ハードウェアを利用可能 パートナー・ソリューションの特集を表示.
