First local agent benchmark for NVIDIA RTX Spark N1X: 27B in 18.7 minutes
ArtificialAnlys · x · 2026-10-16
Artificial Analysis published the first AA-AgentPerf-Local results: an NVIDIA RTX Spark N1X laptop (up to 128GB unified memory, Blackwell GPU) completes its local agent workload on Qwen3.8 27B in 18.7 minutes.
Methodology: the open-source tool replays recorded agent sessions against a local model server — 8 tasks, 168 model turns, up to 56K context, a fixed 31K output tokens per system — measuring per-turn latency. Servers keeping a KV cache only need to read the 450 median new tokens per turn.
The leaderboard covers three models at 4-bit (27B dense to 124B MoE) across five systems. Upcoming additions include IQ4XS and Inco AI's Splash, a Metal engine for Apple Silicon using DFlash speculative decoding, with early tests showing faster completions.
Related event: RTX Spark N1X doubles MacBook Pro in first local agent benchmark(4 posts)→
More from Infra
- Phison CEO calls CXMT memory scare a 'ghost story', warns of materials shortage — BenBajarin · 2026-10-16
- Ben Bajarin: The Future of Compute Is Fungible as Hyperscalers Design for Workload Flexibility — BenBajarin · 2026-10-16
- AirLLM runs frontier-size models layer-by-layer: 671B on 12GB VRAM — Aiden_Tech_Ai · 2026-10-16
- NVIDIA Launches cuPotential in CUDA-X to Bring AI Agents to Semiconductor Materials Simulation — tomaszbednarz · 2026-10-16
- Q4-quantized coder model on 8-core CPU builds interactive solar system demo in one shot — IntrepidMindExplorer · 2026-10-16
- TokenSpeed MLA 2 ships for Vera Rubin: up to 3.44× faster Kimi K3 decode — zhyncs42 · 2026-10-16