howtospark
models / kat-coder-v2-5
KC

35B params~3.5B activebfloat16MoE 256×8GQA 16:2262K ctxmultimodalapache-2.0
Sparks
~80 GB
~63 GB
~45 GB
~28 GB
~21 GB
~21 GB
~20 GB
110 usable
REAP dialno pruning · 256 experts
Spec decode
Context33K / 262KKV
On the Sparks — 1× · 2-bit experts + NVFP4 dense · 33K ctx · FP16 KV~1.1M ctx fits usable
spark
90 GB free
110 usable
20* / 110 GB
18% of usable
Expert planesDense weightsKV cacheActivations + graphsOS reserve

* segment sizes marked with an asterisk are estimates pending a measured run

Estimated per-node footprint, weights + KV sharded across the Spark (tensor parallel). Overhead ~8 GB/node; KV at FP16.

layer map
40 layerslinear attention×30full attention×10
in
out

one MoE layer · 256 experts, 8 fire per token

each expert ≈ 3M params

resident: 40 MoE layers × 256 experts × 3M 32.2B

active/token: 40 × 8 × 3M 1.0B experts

+ 2.4B always-on dense (attention, embeddings, shared stack) = 3.5B active

Eval scores (compare all)

Reported

69.4SWE-bench Verified
63SWE-bench Multilingual
45.96SWE-bench Pro
41.02Terminal-Bench 2.1
93.43PinchBench
44.2SciCode
46.21KAT-Code-Bench

Recipes (all recipes)

Worked deployments that run this model on DGX Spark — each shows how it fills each node's unified memory.

Optimize
Verdict

35B params, ~69 GB at bfloat16 — fits on one Spark as-is.

Make it fit

Get the weights (and headroom for KV) inside the 128 GB pool — the hard gate.

Lossless compression

BF16 weights entropy-code ~30% smaller bit-exact — smaller downloads and less bandwidth per token, no quality question at all.

Make it fast

Once it fits: fewer bytes per token and fewer decode steps against 273 GB/s.

Add a speculative decoder

Decode is bandwidth-bound at ~39.5 tok/s ceiling; a trained EAGLE-style draft multiplies tokens per weight-read.

Prune experts for speed

Fewer experts ⇒ smaller working set ⇒ better cache/page behavior, even when it already fits.

Tame the KV cache

At 262144 tokens the KV cache alone is ~21 GB at FP16; quantize or evict.

Make it yours

Change what the model does — adapt, edit, or steer it — independent of size.

Fine-tune it

LoRA/QLoRA adapts behavior on one Spark without touching the base weights.