Today for AI
HOT RADAR
computeHEAT 0.0°

AI CURATED HOT STORY · 10/15/2026

NVIDIA RTX Spark Laptop Runs Local Qwen3.8 Agent in 18.7 Minutes

6 reports archived0 independent sourcesupdated 10/21/2026, 23:25:18
Synthesis & Latest Updates

NVIDIA has launched the RTX Spark N1X laptop, featuring a Blackwell GPU and Grace CPU with up to 128 GB of unified memory. Initial tests confirm it can run the Qwen3.8 27B model for complex agent tasks locally in just 18.7 minutes, marking a significant leap in consumer-grade hardware's ability to handle large-scale AI inference without cloud dependency.

Cross-verified data shows the device is twice as fast as the MacBook Pro M5 Pro in agent workloads and capable of loading models with weights up to 77 GB. Developers have also successfully deployed Linux on these units with minimal performance loss, expanding their utility for open-source model fine-tuning. Meanwhile, NVIDIA expanded DGX Spark cluster capabilities to support up to 8 devices with 1 TB of unified memory, further boosting local ecosystem power.

This development accelerates the shift of AI applications from cloud to edge. With partners like Lenovo releasing AIPC products starting at $29,999 RMB, the focus on zero-token-cost inference, data privacy, and offline capability promises to democratize local AI creation, coding, and research across mainstream brands.

LATEST/Recent updates show DGX Spark clusters now support up to 8 devices with 1 TB unified memory. Additionally, developers successfully ran Arch Linux on RTX Spark laptops, proving cross-platform viability for local LLMs with less than 1.6% performance overhead.

Hourly heat curve

168 fully observed hours
Now
0.0
Peak
5.311:00
24h change
–

The line compares only sources observed throughout; gap hours are interpolated to keep the trend continuous.

Coverage timeline

Total 6 reports · Latest first
  4. 量子位官网T2·42 pts

    Lenovo Launches YOGA Pro 15 with NVIDIA RTX Spark Chip for Local AI

    • Hardware Breakthrough: Features NVIDIA RTX Spark N1X chip with 6144-core Blackwell GPU and 20-core Grace CPU, delivering 1 Petaflop FP4 performance.
    • Architectural Innovation: Uses up to 128GB unified memory to eliminate traditional VRAM bottlenecks, enabling local execution of massive models.
    • Use Cases: Targets local AI creation, coding, and research, emphasizing data privacy and zero-token-cost inference without network dependency.
  5. Artificial Analysis (@ArtificialAnlys)T1·78 pts

    NVIDIA RTX Spark N1X Laptop Benchmark: Runs Qwen3.8 27B with 128GB Unified Memory in 18.7 Minutes

    • RTX Spark N1X is essentially a laptop version of the GB10 chip, featuring a 6,144-core Blackwell GPU and 300 GB/s memory bandwidth.
    • It runs Qwen3.8 27B agent tasks 2.0x faster than the MacBook Pro M5 Pro, approaching desktop DGX Spark performance.
    • The 128 GB unified memory allows it to load models with 77 GB weights (e.g., Ling 3.0 Flash), surpassing traditional laptop VRAM limits.
  6. Artificial AnalysisT1·68 pts

    NVIDIA RTX Spark N1X First Test: Runs Local Qwen3.8 Agent in 18.7 Minutes with 128GB Unified Memory

    • The NVIDIA RTX Spark N1X laptop features a Blackwell GPU and up to 128 GB of unified memory, optimized for local AI inference.
    • AA-AgentPerf-Local measures hardware performance by replaying recorded agent sessions rather than testing single-generation speed alone.
    • Under this benchmark, the RTX Spark N1X completes a full agent task using the Qwen3.8 27B model in 18.7 minutes.