No: the roughly three-minute WhatsApp task reported by 3DJuegos does not establish that NVIDIA RTX Spark ran the AI locally. The reviewer used Hermes with a ChatGPT subscription to read 50 WhatsApp Web messages and send a reply; the specific model and network path were not stated. Separate LTT Labs tests did run local models, but they measured inference benchmarks—not that messaging workflow. 3DJuegos’ account describes the practical task; LTT Labs’ review supplies the separate test results.
The verdict is mixed: the local-model tests show promise under a narrow benchmark condition, while the gaming comparison is modest and battery endurance changes sharply with workload. The ASUS ProArt P16 makes more sense for someone interested in local inference than for a buyer focused on unplugged gaming. For context, NeoTeo previously covered local-AI use on another RTX Spark laptop; this follow-up centers on the ProArt’s messaging task and separate comparative tests.
What the WhatsApp task actually tells us
In the 3DJuegos account, Hermes handled a WhatsApp Web task involving 50 messages and a reply in about three minutes. The reviewer sometimes had to give it an extra instruction. That is a reported observation from one reviewer’s use of the ASUS laptop, not a measured benchmark of on-device inference.
The distinction matters: the account pairs Hermes with a ChatGPT subscription, while leaving the model and network route unspecified. LTT Labs’ local-model results are a different workload and cannot identify how the WhatsApp task ran.
Local inference was tested separately
LTT Labs used llama.cpp to test Gemma 4 E4B Q8_0 and, on systems with enough memory, Qwen 3.8 27B Q4_K_M. Its benchmark used context depth 0, a best-case condition that does not reproduce an agent handling a long conversation. In the Gemma test, LTT Labs found N1X systems strong at prompt processing—the stage that reads the input—while token generation, or producing the response, was broadly similar across several tested platforms. Model choice, inference engine, quantization and speculative decoding can all affect the result.
That makes the test useful for understanding local inference, but it does not turn the WhatsApp example into proof of local execution. The two results answer different questions.
Gaming looked adequate in rasterization tests
The 3DJuegos reviewer described smooth play in Resident Evil Requiem with high settings, ray tracing and DLSS, and also played Gears of War: E-Day at high settings. That is a subjective report. LTT Labs’ four-game comparison instead measured native rasterization, without upscaling or frame generation: an Alienware system with an RTX 5070 Ti led, while N1X performance was adequate rather than outstanding.
For someone prioritizing gaming, the RTX 5070 Ti system is the more persuasive performance reference in that test. The two accounts used different conditions, so the hands-on impression and native-rasterization result are not interchangeable measures.
ProArt battery life swings by workload
The ASUS ProArt P16’s reported endurance varies widely across gaming, video playback and graphics stress. These figures come from different workloads and test methods:
| Workload | Reported endurance | Conditions |
| Gaming | About 2 hours | 3DJuegos reviewer’s use from a full charge |
| Local video playback | 12 hours 2 minutes | LTT Labs, local 1080p video |
| Graphics stress | 1 hour 9 minutes | LTT Labs, 1440p GravityMark at maximum-performance settings |
ASUS lists a 99.9 Wh battery for the ProArt configuration. In its graphics stress test, LTT Labs recorded staged reductions in system power as the charge fell. A light video workload and a maximum-performance graphics test therefore describe very different unplugged experiences.
The tested ProArt configuration and reported price
LTT Labs tested an ASUS ProArt P16 H7607B with 64 GB of memory. ASUS’s U.S. listing for the ProArt P16 H7607 specifies a 20-core NVIDIA Grace CPU, a 6,144-core Blackwell RTX GPU, 64 GB of LPDDR5X unified memory and a 1 TB PCIe 4.0 SSD. The listed display is a 16-inch OLED touchscreen with a 2,880 × 1,800 resolution and a 120 Hz refresh rate. ASUS’s product page gives the configuration details; NVIDIA describes up to 128 GB of unified memory across the RTX Spark platform, not as the specification for every laptop. NVIDIA’s platform announcement sets out that platform-wide ceiling.
LTT Labs reported a price of $4,399.99 for its 64 GB ASUS ProArt P16 test configuration.
