sm121

Qwen3.5-122B-A10B on DGX Spark: 28.3 → 51 tok/s (+80%)

Measured SM121 compatibility recipe for MiniMax H3 FL2VA on one NVIDIA DGX Spark with vLLM-Omni and online FP8.

DGX Spark research and tests - containers, benchmarks, and investigation notes for running models on GB10 (SM 12.1)

7.67× LoRA / 8.35× Full FT speedup for Qwen3.5 (0.8B–27B) on NVIDIA DGX Spark — wall-clock parity with rented H100. Lossless within BF16. Three-command interactive wizard handles model picker, data validator, training, and merge.

DeepSeek-V4-Flash DSpark speculative decoding on 2x DGX Spark (GB10/sm_121), 1M context — our fp8 recipe + independent reproduction and cross-build benchmarks of the NVFP4-KV build. Honest, apples-to-apples measurements.

Qwen3.8-Flash-Next as Cogni-Brain on NVIDIA DGX Spark (GB10): HashK GPU PLE + SGLang NEXTN, 36.8 tok/s code, 100/100 tool-eval, 262K context.

Measured serving recipes for DeepSeek-V4.1-Flash on 4x NVIDIA DGX Spark (GB10): 1M context on vLLM (CUDA graphs, vision, tools, DSpark) and a switchless-ring SGLang TP4 lane, plus a cross-project reference table. EN + 中文.

Patches + recipe to deploy festr2/MiMo-V2.5-Pro-NVFP4-MXFP8-attn-TP8 on 8-node DGX Spark sm_121 (Ray + vLLM, TP=8). Fixes the fused-qkv loader bug that mis-slotted Q values as K/V on 7 of 8 ranks.

SM121-optimized vLLM serving of nvidia/GLM-5.3-Flash-NVFP4 on 2x GB10 (TP=2, DFlash2 K=7): patches, click-run container, serve profiles, sanitized evidence

Ornith-1.5-35B NVFP4-W4A16 + BF16 MTP via SGLang on SM12X (SM120/SM121). Click-run serve + SM12X-LLM-BENCH claim row.

Measured vLLM serving for 1–2 NVIDIA DGX Spark (GB10): TP=2 RoCE flagship, validation ledger, stranger-ready Path A/B onboarding.

MiMo-V2.6-Flash-RL NVFP4 on 2x GB10/SM121. Preferred serve is SGLang EAGLE MTP. MIT.

Measured GLM-5.2-Vision NVFP4 reproduction on 8x DGX Spark: 1M context, native SM121, c1/c8 throughput, and fresh 60-minute stability.

Complete GB10 (DGX Spark, sm_121) block-FP8 Triton kernel configs for DeepSeek-V4-Flash-0731 at TP=2, including the DSpark main_proj shape missing from every earlier run

GLM-5.3-Flash (320B MoE, EXL3 4bpw) on 2x NVIDIA DGX Spark (GB10) with vLLM, TP=2: single-stream decode 45-99 tok/s by workload, ~3,000 tok/s prefill (32k prompt), 2M-token fp8 KV cache. Optimization log, raw benchmarks, CUDA kernels, wrong-measurement notes, reproduction steps. AGPL-3.0.

One-click MiniMax-H3 (ComfyUI) deployment tuned for NVIDIA DGX Spark GB10/SM121. Maintained fork: 2026-09 upgrade guide (ComfyUI v0.37 + Spectrum v0.2.28 + KJNodes v1.5.2), int8 black-video workaround, and Turbo LoRA 8-step workflow (1.8x speedup, measured).

Empirical NVIDIA GB10 and SM121 microarchitecture characterization