MiaAI Lab (@MiaAI_lab) added a TP=3 path to GLM-5.3-Flash EXL3 for three DGX Sparks: +28% decode and +9% prefill vs TP=2, with 3.2M KV cache and 1M context, plus optional NFS weight sharing from the head node (NFS_SHARE=1). Repo: github.com/MiaAI-Lab/GLM-5.3-Flash-EXL3-2x-DGX-Sparks.

Key Takeaways

  • ✓New TP=3 path runs GLM-5.3-Flash EXL3 across three DGX Sparks.
  • ✓Vs TP=2: +28% decode, +9% prefill; 3.2M KV cache and 1M context.
  • ✓Optional NFS_SHARE=1 loads weights from the head node to free other Sparks.
⚡

Key Decision Metrics at a Glance

TL;DR Verdict
🏆 Benchmark RankRank
Vs TP=2: +28% decode, +9% prefill; 3.2M KV cache and 1M context.
Full Leaderboard→
💰 Cost & PlansCompare
GLM-5.3-Flash · in: $0.15 / out: $0.5
Compare Plans↗
🛠️ Ecosystem & IDERules
GLM-5.3-Flash · EXL3 · DGX Spark
Setup IDE Rules↗
🧭

Turn your technical choice into a development budget

Compare 40 dev plans & simulate token costs vs $20/mo subscriptions

⚔️ Compare Model→💰 Compare plans & limits↗
Action HubReady to adopt this in production?

Benchmark side-by-side against alternatives, or calculate monthly token cost vs subscription break-even.

⚔️ Launch VS Compare→💳 View Plans↗
Up Nextllama.cpp v0.6.0 ships GLM-5.3-Flash (320B) and Qwen4Exp MTP speculative decoding, a new llama_batch_ext API, and up to ~3x faster few-row mat-mul on Apple GPUs
Read Now→