# GLM-5.2 (unpruned) at 800K max context on 4× DGX Spark 20-33 tok/s — NF3-hybrid checkpoint **URL:** \ **Category:** DGX Spark / GB10 Projects\ **Created:** [July 7, 2026, 12:12pm UTC](https://forums.developer.nvidia.com/t/glm-5-2-unpruned-at-800k-max-context-on-4x-dgx-spark-20-33-tok-s-nf3-hybrid-checkpoint/375909 "2026-07-07T12:12:52Z")\ **Posts on this page:** 3\ **Page:** 1
**Author:** ![tonyd615](https://sea2.discourse-cdn.com/nvidia/user_avatar/forums.developer.nvidia.com/tonyd615/32/511662_2.png) [@tonyd615](https://forums.developer.nvidia.com/u/tonyd615)\ **Post date:** [July 7, 2026, 12:12pm UTC](https://forums.developer.nvidia.com/t/glm-5-2-unpruned-at-800k-max-context-on-4x-dgx-spark-20-33-tok-s-nf3-hybrid-checkpoint/375909/1 "2026-07-07T12:12:52Z")
So me and FABLE had a long night to some success off the shoulders of @CosmicRaisins and @Zatz hard work. We swapped models to get our KV POOL to 875k. I feel like it seems a little slower than what I was getting with the original 200k context and the 600k context version. Maybe there is some tuning but I wanted to atleast stand this up so if anyone wants to give it a shot they can getting speeds up. We run a 4× DGX Spark cluster (GB10, 200GbE fabric) and have been following the GLM-5.2 work in [the 4x GB10 cluster thread](https://forums.developer.nvidia.com/t/glm-5-2-on-a-4x-gb10-cluster-22-tok-s-decode-256k-ctx-recipe/374125) — everything below builds directly on the recipes from that thread. Recently madeby561 published [GLM-5.2-MXFP8-NVFP4-NF3-Hybrid](https://huggingface.co/madeby561/GLM-5.2-MXFP8-NVFP4-NF3-Hybrid): the unpruned model (all 256 experts) hybrid-quantized to **327GB** — top-64 experts per layer in NVFP4, the rest in a custom 3-bit “NF3” format, MXFP8 elsewhere. It targets 4× RTX PRO 6000 (sm\_120/amd64), but since it’s ~78GB smaller than the Int4-Int8Mix checkpoint most of us run, we wondered what the freed memory would do for KV on Sparks. So we pointed the existing community recipes at it — b12x/eldritch stack rebuilt for sm\_121a/aarch64, TP4 + DCP4 `ag_rs` with Zatz’s 640K launch settings as the template, MTP speculative decode. Result: - **KV pool: 877,056 tokens** , currently serving at **`--max-model-len 800000`** — clean boot, coherent output, tool calling working - Decode at 800K max-len matches a 240K-max-len boot at equal fill (shallow-context bench; we haven’t run a depth sweep yet, so no claims about decode speed at 700K fill) Since DCP trades speed for context (as CosmicRaisins’ DCP2 results showed), the repo keeps two launch shapes: **DCP4 “context” (800K max-len, 877K pool):** c1 19.9 / c2 32.5 / c3 35.3 / c4 44.1 / c6 38.7 aggregate tok/s (512-tok gens, temp 0, MTP k=4, acceptance 3.2–3.7) **DCP1 “fast” (200K max-len, 219K pool):** c1 24–29 (content-dependent, acceptance 3.3–4.0) / c4 54.2 / **c6 67.4 aggregate** Port notes for anyone trying this checkpoint on GB10 (it needs madeby561’s custom serving stack): the NF3 CuteDSL kernel runs on sm\_121a essentially as-is — it even ships an SM121 register-count table — but three things bit us: (1) the hybrid loader hardcodes its `mxfp8_tier.json` path to the author’s image layout (use the `HYBRID_MXFP8_TIER_JSON` env override), (2) warmup needs the patched `b12x_sparse_indexer_warmup.py` from the author’s docker image, which isn’t in any public repo yet, and (3) eugr-style aarch64 builds need `nvidia-cutlass-dsl==4.5.2` plus its `.pth` path hook added. Full replication recipe, both launch envs, and all fixes: [GitHub - tonyd2wild/GLM-5.2-NF3-Hybrid-4x-DGX-Spark · GitHub](https://github.com/tonyd2wild/GLM-5.2-NF3-Hybrid-4x-DGX-Spark) Credits: **madeby561** for the NF3 format, kernel, and checkpoint — this is his work, we just ported it; **lukealonso** for b12x and the NVFP4 donor tier; **Zatz** for the DCP recipe our launch config is copied from; **CosmicRaisins** for the stack that got GLM-5.2 running on these boxes in the first place; **m9e/eugr** for the build harness; **aidendle94** for pointing at the checkpoint. Happy to share anything else that’s useful, and very open to tuning suggestions — single-stream on the DCP4 shape (19.9) feels like it has headroom. ---
**Author:** ![p33zy](https://developer.download.nvidia.com/images/forums/profile-default-devtalk-84.png) [@p33zy](https://forums.developer.nvidia.com/u/p33zy)\ **Post date:** [July 8, 2026, 10:32am UTC](https://forums.developer.nvidia.com/t/glm-5-2-unpruned-at-800k-max-context-on-4x-dgx-spark-20-33-tok-s-nf3-hybrid-checkpoint/375909/2 "2026-07-08T10:32:05Z")
Awesome work! What’s the prefill speed like on this? Any improvements over the other recipe or is it slower? ---
**Author:** ![p33zy](https://developer.download.nvidia.com/images/forums/profile-default-devtalk-84.png) [@p33zy](https://forums.developer.nvidia.com/u/p33zy)\ **Post date:** [July 12, 2026, 5:30pm UTC](https://forums.developer.nvidia.com/t/glm-5-2-unpruned-at-800k-max-context-on-4x-dgx-spark-20-33-tok-s-nf3-hybrid-checkpoint/375909/3 "2026-07-12T17:30:22Z")
Got this up and running - DC1 prefill was around 800 t/s.