Mac Studio M5 Ultra vs Dual DGX Spark: Long-Input Inference Benchmarks Show Split Verdict
GabGarrett · x · 2026-10-19
Testing at a fixed 200K context cap with inputs scaled from 30K to 190K tokens, satsuki2694 benchmarked a Mac Studio M5 Ultra against dual DGX Sparks running Qwen, DeepSeek V4 Flash 0731, and GLM locally.
Key findings:
- Qwen: Mac wins on TTFT, but Spark×2 holds decode speed much better at long inputs (Mac prefill 4700 tok/s falling vs Spark steady 2000).
- DeepSeek V4 Flash: Spark×2 pulls ahead clearly on long inputs — Mac decode drops from 60 to 25 tok/s with steeply rising TTFT, while Spark's automatic KV management keeps decode flat.
- GLM Flash: near-parity on decode and TTFT; Spark's prefill stays flat while Mac declines slightly.
Bottom line: Mac gives faster first tokens at short inputs; dual DGX Sparks are more stable for long-context decode.
More from Infra
- Sipeed's MaixStation challenges DGX Spark: 128GB LPDDR5X at 40W, one-third the size — AIFlow_ML · 2026-10-19
- ReaDy Uses Recurrent States as Dynamic Context for Speculative Decoding in Hybrid Linear Attention LLMs — Bohan Sun · 2026-10-19
- TRELLIS.2 image-to-3D running on AMD RX 9060 XT with ROCm: fixes, workflows, MIT swaps — ova-mojica · 2026-10-19
- Celesto Claims CI Runner With 96% Lower Per-Minute Cost Than GitHub Actions — aniketmaurya · 2026-10-19
- Huawei's Atlas 960E superpod links 4,096 NPUs with near-packaged optics — bookwormengr · 2026-10-19
- Robot brain inference cost fell 84% in two months, from $2.36 to $0.38 per task — victor_explore · 2026-10-19