GLM-5.3-Flash (320B MoE) at TP4 across four DGX Sparks, same day as the model drop: 36 tok/s, 1.26M-token FP8 KV pool, 262K context, MTP spec decode. First TP4 glm5_next outside B200 hardware. Full patched-image recipe + the GB10 memory study.
☆57Oct 10, 2026Updated last week
Alternatives and similar repositories for GLM-5.3-Flash-NVFP4-1M-KV-4x-DGX-Spark
Users that are interested in GLM-5.3-Flash-NVFP4-1M-KV-4x-DGX-Spark are comparing it to the libraries listed below. We may earn a commission when you buy through links labeled 'Ad' on this page.
Sorting:
- GLM-5.3-Flash NVFP4 on 4x NVIDIA DGX Spark: vLLM TP4, DFlash2 adaptive draft, measured FP8 vs NVFP4 vs EXL3 vs SGLang☆37Sep 29, 2026Updated 2 weeks ago
- RTX 6000 96GB/Spark 128GB vLLM Recipe for Qwen 3.8 Flash Next☆22Sep 8, 2026Updated last month
- GLM-5.3-Flash (NVFP4) at TP4 across 4x DGX Spark via a switchless RoCE ring + DFlash2 — reproducible recipe☆108Updated this week
- GLM-5.3-Flash EXL3 on 2x DGX Spark with TensorFold☆457Updated this week
- GLM-5.3-Flash (NVFP4) on 2x NVIDIA DGX Spark - vLLM TP2, 262K context, MTP. World-first deploy recipe: 7 day-0 bugs found and fixed, patc…☆236Sep 30, 2026Updated 2 weeks ago
- Deploy to Railway using AI coding agents - Free Credits Offer • AdUse Claude Code, Codex, OpenCode, and more. Autonomous software development now has the infrastructure to match with Railway.
- Local image generation and editing for NVIDIA DGX Spark, powered by Qwen-Image-2.1 and Gradio.☆31Sep 21, 2026Updated 3 weeks ago
- Switchless RoCE collectives, NCCL profiles and fabric tooling for DGX Spark inference recipes☆37Updated this week
- CUDA Blackwell performance fork of antirez/ds4 (DeepSeek-V4-Flash): ~4x prefill, ~1.5x decode, continuous batching, weight server, DSpark☆77Sep 4, 2026Updated last month
- Plug-and-play, dependency-free web command center for DGX Spark fleets (and any GPU boxes): live per-node GPU/CPU/mem/temps/power, option…☆48Updated this week
- GLM-5.3-Flash EXL3 (320B MoE) on 2x NVIDIA DGX Spark — production serving kit, 1M context, 97%+ multi-session prefix caching, DFlash2 spe…☆62Oct 5, 2026Updated 2 weeks ago
- GLM-5.3 (full) on 3-6 DGX Sparks (GB10): a from-scratch CUDA inference engine☆37Updated this week
- Minimal read-only OS image for GB10 (DGX Spark / ASUS GX10) boxes☆89Oct 4, 2026Updated 2 weeks ago
- Capping GPU clocks on NVIDIA GB10: -12C peak temp, -36% GPU power, ~1.6% throughput cost. Measurement harness, raw data, and the systemd …☆77Jul 31, 2026Updated 2 months ago
- GLM-5.3 Flash EXL3 for 2-4x DGX Sparks☆736Updated this week
- Managed Kubernetes at scale on DigitalOcean • AdDigitalOcean Kubernetes includes the control plane, bandwidth allowance, container registry, automatic updates, and more for free.
- GLM-5.2-Int4-Int8 on 8x GB10 cluster☆39Jul 30, 2026Updated 2 months ago
- keys-DGX-Sparkticus Utimate Power Pack Unleashed (Dual DGX-Sparks Needed)☆43Aug 23, 2026Updated last month
- Qwen3.8-Flash-Next-NVFP4-vLLM-Dual-DGX-Spark☆411Sep 29, 2026Updated 2 weeks ago
- DeepSeek V4.1 Flash on 4x RTX PRO 6000 Blackwell: native weights, DSpark, vision, NVMe or locked RAM Engram offload.☆73Updated this week
- vLLM inference solutions for NVIDIA GB10 (DGX Spark, sm121) — KIMI-K3 B12X_MLA + DSpark, GLM-5.2☆67Sep 30, 2026Updated 2 weeks ago
- 2x DGX Spark recipe for LibertAIDAI/GLM-5.3-Flash-NVFP4 (NVFP4, Ray TP=2, multimodal)☆90Sep 1, 2026Updated last month
- One command turns a DGX Spark or ASUS Ascent GX10 into a private AI server: seven measured, pinned AI lanes (text up to 1M tokens, images…☆229Updated this week
- Pinned one-DGX-Spark Docker recipe for DeepSeek V4 Flash with EXL3, SparkInfer, and 262K NVFP4 MLA KV cache☆166Sep 3, 2026Updated last month
- Qwen3.8-27B on one DGX Spark: SGLang + DFlash2 recipe, engine image, benchmarks, data and charts. By @redp314☆37Sep 6, 2026Updated last month
- Wordpress hosting with auto-scaling - Free Trial Offer • AdFully Managed hosting for WordPress and WooCommerce businesses that need reliable, auto-scalable performance. Cloudways SafeUpdates now available.
- GLM-5.2 NVFP4+AQLM on 3× DGX Spark — 380k context MTP serve stack☆72Aug 2, 2026Updated 2 months ago
- Run DeepSeek-V4-Flash at full 1M context AND two MiniMax H3 video instances on the same two DGX Sparks. Benchmarked C1-C6.☆35Aug 7, 2026Updated 2 months ago
- Qwen3.8 27B for DGX Spark / RTX 6000 PRO☆100Aug 14, 2026Updated 2 months ago
- MiniMax-H3 on single DGX Spark: keys Sol-Engine ports + Heretic TE + Spectrum/Sage speed + 0.5MP upscaler finish. Formal bench included.☆49Aug 23, 2026Updated last month
- sparkDash ⚡ — Multi-DGX Spark Monitoring Dashboard☆715Updated this week
- Decision engine☆51Sep 24, 2026Updated 3 weeks ago
- DeepSeek-V4.1-Flash on 3-4x NVIDIA DGX Sparks☆247Updated this week
- Optimized SGLang runtime for Qwen3.8-27B FP8 with DFlash2 and Qwen3.8 Flash-Next NVFP4 with FR-Spec on one NVIDIA RTX PRO 6000 Blackwell …☆167Updated this week
- LLM Inference Engine for Metal and CUDA.☆1,375Updated this week
- Deploy to Railway using AI coding agents - Free Credits Offer • AdUse Claude Code, Codex, OpenCode, and more. Autonomous software development now has the infrastructure to match with Railway.
- A collection of AI agent skills for QEMU hardware modeling☆58Jul 28, 2026Updated 2 months ago
- Validated GLM-5.3 Flash recipe for 2x NVIDIA RTX PRO 6000 Blackwell 96GB: 262K context, EXL3/TR3, adaptive MTP, tools, and vision.☆58Sep 20, 2026Updated 3 weeks ago
- Strata fork: Qwen3.8-Flash-Next 125B MoE in NVFP4 on one RTX 20-50 card (12 GB+) and 64 GB of RAM or more. Setup installs our GPTQ quants…☆165Updated this week
- Low-latency collective transport and distributed LLM serving stack for switchless GB10 clusters.☆94Updated this week
- English guide + one-command patcher for the vLLM spin-wait bug that overheats NVIDIA DGX Spark / GB10. Credit: nacyot. CPU cores spin at …☆24Aug 17, 2026Updated 2 months ago
- Qwen3.8-27B on SGLang for NVIDIA RTX PRO 6000 (96GB, SM120): DSpark speculative decoding, full 256K context, 8 concurrent requests☆60Aug 24, 2026Updated last month
- Heterogeneous prefill/decode for DeepSeek-V4-Flash: CUDA prefill (DGX Spark, vLLM) -> Metal decode (Mac Studio, oMLX) over plain 10GbE☆151Updated this week