---

base_model: Motif-Technologies/Motif-3

base_model_relation: quantized

license: mit

language:

- en

- ko

pipeline_tag: text-generation

library_name: gguf

tags:

- gguf

- motif

- motif-3

- mixture-of-experts

- mixed-quantization

- iq2_xxs

- dgx-spark

- long-context

---

Motif 3 — Mixed-Quant GGUF

**A structurally intact 314.8B-parameter Motif-3 MoE compressed to 94.16 GB

(87.70 GiB).

Native H200 residency, latent-KV execution, OpenAI serving, tools, continuous

batching, and retrieval through 128K are validated. The ds4-dfm path also

passes a strict 262,080-token prompt plus decode gate on one DGX Spark.**

Support my work

I work on making large language models practical on hardware they were never really designed to fit on — through mixed quantization, inference optimization, custom kernels, and serving experiments.

While much of the development happens on local hardware, calibration, profiling, and large-scale validation often require expensive on-demand GPUs.

Contributions help pay for that compute, storage, and testing infrastructure, so I can keep experimenting and publishing the results openly.

<a href="https://www.buymeacoffee.com/baekpica" target="_blank"><img src="https://cdn.buymeacoffee.com/buttons/v2/default-yellow.png" alt="Buy Me a Coffee" style="height: 60px !important;width: 217px !important;"></a> <a href="https://github.com/sponsors/Baekpica" target="_blank"><img src="https://img.shields.io/badge/Sponsor-EA4AAA?style=for-the-badge&logo=githubsponsors&logoColor=white" alt="Sponsor Baekpica on GitHub" style="height: 60px !important;width: 217px !important;"></a>

Full-topology mixed-quant conversion of

Motif-Technologies/Motif-3,

designed as the resident-weight baseline for one 128 GB-class DGX Spark.

This is an independent conversion, not an official Motif Technologies release.

Nothing is pruned, merged, expert-dropped, or layer-dropped. The artifact

retains all 53 target-model layers, the first two dense layers, all 51 sparse

layers, all 384 routed experts per sparse layer with top-8 routing, the shared

expert, Grouped Differential Latent Attention (GDLA), Expert-Specific

PolyNorm, modified mHC, and the complete one-layer MTP predictor.

| | |

|---|---:|

| Architecture | Motif-3, 53 layers, 14 full + 39 SWA GDLA |

| Routed experts retained | 384 / 384 per sparse layer |

| Tensors | 2,287 |

| Artifact | 94.16 GB (87.6957 GiB), 11 shards |

| Native H200 model/runtime | 91.262 GiB, conservative measured repeat |

| Native 256K session | 4.037 GiB, measured |

| Source context limit | 262,144 tokens |

| H200 semantic execution | short/32K/64K/128K passed; 256K partial prefill only |

| Single-GB10 execution | 262,080-token prompt + 43-token decode; three-way 196K-context serving passed |

ds4-dfm

This model is served by ds4-dfm, a common ds4 release line for

DFM (독자 파운데이션 모델, 독파모) model families. It was developed to

serve full-scale models on a single NVIDIA DGX Spark with 128 GB of unified

memory through explicit model-family C/CUDA paths and device kernels.

The release line is the

dfm branch of

Baekpica/ds4. The integration tag is

v0.6.2-dfm

(Entrpi v0.6.2 absorbed). The 32K/256K HTTP gates below remain the earlier

Spark evidence at

593d251;

the original 8K ds4-bench point uses

cc2f277;

the persistent multi-bank measurements use

cf605e0.

The 8K/32K remesure uses

a09ff4f on dfm.

The 256K remesure uses

6500480

(kernels through a09ff4f). The integration tag is not moved. One

ds4-server command and one HTTP contract cover the integrated DeepSeek,

Solar Open2, K-EXAONE, Motif-3, and dots3-note families; only the GGUF path

and its matching weight-owner manifest change. The server provides OpenAI

Chat Completions, OpenAI Completions, OpenAI Responses, and Anthropic

Messages.

The production GGUF in this repository was loaded through the common release

binary on DGX Spark. Context and throughput claims are limited to the exact

conditions below; model quality has not been evaluated by these serving tests.

Artifact

| Variant | Split | Exact size | Purpose |

|---|---:|---:|---|

| MQ87-88-FIT | 11 files | 94,162,542,816 bytes (94.16 GB / 87.6957 GiB) | one-Spark capacity/release baseline |

The canonical merged/unsharded byte stream used by native ds4 is

94,162,541,472 bytes (94.16 GB / 87.6957 GiB) with SHA-256

15755a735753bc1396e5ffa539e65a779a4fd769e8833360a4d743c4c60c2f25.

Start with Motif-3-MQ87-88-FIT-00001-of-00011.gguf; compatible split-aware

runtimes discover the remaining shards automatically. Exact per-shard hashes

are published in MQ87-88-FIT-SHA256SUMS.

The weight files are fixed at Hub revision

efd6044e25e7f8e3b459a737d021091e2e69b6c6. Later model-card/report commits

do not change shard bytes or hashes.

hf download Baekpica/Motif-3-Mixed-Quant-GGUF \
  --revision efd6044e25e7f8e3b459a737d021091e2e69b6c6 \
  --include 'Motif-3-MQ87-88-FIT-*.gguf' \
  --include MQ87-88-FIT-SHA256SUMS \
  --local-dir ./Motif-3-MQ87-88-FIT

The name is a nominal recipe label, not the artifact's decimal size. The

actual payload is 94.16 GB (87.70 GiB), and the exact byte counts above are

authoritative. Admission depends on whether its measured aligned/repacked

resident form fits the target Spark together with latent KV, workspace, and

server state.

Precision recipe

Precision is assigned by module role. Always-active and control paths stay

conservative; the 51×384 routed-expert stack carries the compression burden.

| Tensor group | Type | Rationale |

|---|---|---|

| Token embedding and LM head | Q8_0 | token/logit fidelity |

| GDLA projections and differential/output-gate paths | Q8_0 | attention and long-context stability |

| Dense MLP, shared expert, MTP projections | Q8_0 | always active |

| Routed expert gate/up, layers 2–52 | IQ2_XXS + Q8 imatrix | dominant parameter mass |

| Routed expert down, layers 2–52 | Q2_K + Q8 imatrix | capacity baseline |

| Router weights | F32 | top-8 decision stability |

| RMSNorm, Expert PolyNorm coefficients/biases | F32 | normalization/activation stability |

| mHC controls/scalars | F32 | sigmoid, clamp, Sinkhorn stability |

| mHC small projections | BF16 | small protected matrices |

The completed GGUF contains 2,287 tensors: **1,273 F32, 318 BF16, 543

Q8_0, 102 IQ2_XXS, and 51 Q2_K**. The 102 IQ2 tensors are separately stored

gate and up matrices. They originate from 51 fused checkpoint gate/up tensors;

the split is lossless before quantization and retains every expert value.

No Q4 edge-layer promotion is included in this release. MQ95/MQ97 are outside

the scope of this capacity-first artifact.

Q8_0 activation calibration

Calibration was collected only after producing and freezing the separate

full-topology Q8_0 reference in

Baekpica/Motif-3-GGUF.

The corpus reuses the checksum-pinned normalization and target shares audited

for the preceding Solar Open 2 mixed-quant release, but every record is

rendered and counted again with Motif-3's official final tokenizer and chat

template. It contains 1,539 documents, 4,079,555 official-tokenizer tokens,

and 17,011,959 bytes.

| Calibration bucket | Target token share |

|---|---:|

| Instruction-following chat | 22% |

| Cascade stage 1 reasoning | 16% |

| Cascade stage 2 reasoning/tools | 16% |

| Korean | 16% |

| Other multilingual | 12% |

| Finance | 6% |

| SWE agentic | 6% |

| Algorithmic code | 6% |

The H200 calibration runtime directly dequantized the Q8_0 GGUF and executed

the official Motif-3 equations with a layer-major schedule: mHC, expanded

historical K/V GDLA, interleaved 128-token SWA/full attention, YaRN, the

differential signal/noise heads, elementwise attention output gate, sigmoid

top-8 routing with correction-bias selection and route renormalization, shared

experts, and per-expert PolyNorm.

| | |

|---|---:|

| Calibration chunks | 590 × 512 tokens |

| Total official-tokenizer tokens | 302,080 |

| Routed observations | 123,248,640 |

| Sparse layer/expert cells | 51 × 384 |

| Zero-coverage cells | 0 |

| Routes per layer/expert | min 313, median 5,934, p95 9,986.7, max 85,011 |

| Imatrix size | 742,004,501 bytes |

| Imatrix SHA-256 | 54fcc4d6d1fe96a3fc12bd24869eff128c724737ac07ce866ce08018a5e3cfbc |

Gate/up importance observes the exact FFN-normalized input. Down importance

observes the Expert-Specific PolyNorm output after route weighting, matching

the ds4 routed-MoE execution order. No uniform, random, proxy-model, or

source-BF16 activation matrix was substituted.

The machine-readable activation report is published here as

Q8_0-IMATRIX-REPORT.json. Corpus construction is included in the public

reproduction materials; the exact rendered corpus, final imatrix, and

rank-local accumulators are preserved in the private Spark handoff.

Provenance

| | |

|---|---|

| Source model | Motif-Technologies/Motif-3 |

| Exact source revision | ccceb1a5fd7b5eb32e47841216b3caf5666c07bc |

| Source parameters | 314,841,775,750 |

| Source tensors | 2,236 |

| GGUF tensors | 2,287 |

| Native context metadata | 262,144 tokens |

| Full Q8_0 reference | Baekpica/Motif-3-GGUF@5c266c95bf8c8d822d50e5e1cce9d108eaadb2af |

| Fixed mixed-weight revision | efd6044e25e7f8e3b459a737d021091e2e69b6c6 |

| Official implementation oracle | MotifTechnologies/vllm@4cd9eb4129883565e69d508038d783d59ee01867 |

| Conversion base | ggml-org/llama.cpp@1d2869c6e54d5003f3927a79efbca0fefa034a6d |

| ds4 base | Baekpica/ds4@b0309611041655f4e45671cfd9c9886aff161406 |

| Native ds4 implementation | Baekpica/ds4:feature/motif-3-model-loader@d878ea1a1d67bc0f0bd60e20e75b4a011aa2d8d9 |

| DGX Spark long-context implementation | Baekpica/ds4@593d251 |

| DGX Spark 8K throughput implementation | Baekpica/ds4@cc2f277 |

| DGX Spark multi-bank implementation | Baekpica/ds4@cf605e0 |

| DGX Spark 8K/32K remesure (dfm) | Baekpica/ds4@a09ff4f |

| ds4-dfm integration tag | v0.6.2-dfm |

| Public reproduction | Baekpica/motif-3-mixed-ds4 |

| Private Spark handoff | Expensive calibration state plus offline reproduction/runtime snapshots are preserved in hf://buckets/Baekpica/motif-3-spark-handoff |

Only the official final Motif-3 checkpoint was used. Motif-3-Beta was not used

as a source, calibration input, implementation oracle, or fallback.

Model structure and context memory

Layers whose index is divisible by four use full attention (14 layers); the

other 39 layers use a bounded 128-token sliding window. The production target

does not retain expanded historical K/V. Its persistent context state is

latent KV plus the decoupled RoPE key for full-attention layers, bounded SWA

ring state, and cache identity/position state.

On H200, creating a native 262,144-token session while the model remained

resident allocated 4,236,751,872 bytes (3.946 GiB) of cache tensor payload and

produced a 4,334,813,184-byte (4.037 GiB) CUDA free-memory delta including

allocator overhead. This is a physical H200 measurement, not a Spark claim.

Physical unified-memory residency and OS headroom on the target GB10 remain

authoritative.

The higher of two native-sm_90 H200 model/runtime initialization repeats plus

the 256K-session delta was 102,326,337,536 bytes (95.298828125 GiB). This supports the capacity design;

it does not predict GB10 driver, allocator, or OS overhead.

The conservative H200 component measurements place model/runtime at 91.262

GiB and the native 256K session at 4.037 GiB, inside the design preferences of

98 GiB and 5 GiB respectively; only the target GB10 can establish unified-

memory admission and final MemAvailable.

Runtime compatibility

Motif 3 is not a Llama-family graph. A compatible runtime must implement its

384E sigmoid router, route normalization/scale, shared expert,

Expert-Specific PolyNorm, modified mHC, GDLA/differential heads and output

gate, interleaved SWA/full attention, YaRN, latent KV semantics, and MTP.

The target runtime is

Baekpica/ds4 at the exact ds4-dfm Spark

commit pinned above. The earlier feature/motif-3-model-loader revision is

retained as H200 bring-up provenance. The private Spark handoff also carries

an offline source snapshot and commit metadata.

Stock GGUF runtimes should not be assumed to execute this architecture merely

because they can parse the container.

The current native branch has an explicit Motif tensor binder and CUDA graph,

production latent-KV/SWA-ring sessions, strict device-resident model loading,

the official tokenizer/chat/reasoning/tool protocol, and an OpenAI-compatible

ds4-server path. Motif sessions refuse streaming/offloaded weights and never

fall through to the generic DeepSeek graph.

A clean build at the pinned Spark revision completed the CPU and cuda-spark

targets, linked the five runtime programs, and produced only sm_121a CUDA

code objects. The same revision passed the Motif loader, tokenizer, CUDA

primitive, real-weight parity, resident-memory, and strict OpenAI gates below.

The public files use standard GGUF splitting. The current ds4 development

loader consumes one merged GGUF, so merge from the first shard with

llama-gguf-split --merge before launch. This is a one-time artifact assembly,

not SSD weight streaming; production admission still requires the merged or

repacked weights to be resident and forbids simultaneous physical residency of

duplicate raw/repacked mappings.

Validation status

The unsharded artifact and the complete 11-file split set independently passed

strict source-map validation. The verifier checked the pinned revision,

architecture metadata, native 262,144-token context, all 53 target layers,

all routed gate/up/down tensors for sparse layers 2–52, all 384 experts, the

shared experts, and the complete MTP block. It found exactly 2,287 unique

tensors with no missing, duplicate, unexpected, mistyped, misshaped, or

out-of-bounds payloads. The split set totals 94,162,542,816 bytes (94.16 GB /

87.6957 GiB); every shard also has a published SHA-256 digest.

A separate 57-row numerical comparison sampled the embedding and LM head,

GDLA/control paths, dense/shared paths, routed experts at layers 2, 26, and 52,

both halves of the checkpoint's fused gate/up weights, and MTP. Protected

F32/BF16 rows were source-exact. Minimum sampled cosine was 0.9999740 for

Q8_0, 0.9417932 for IQ2_XXS, and 0.9580462 for Q2_K. The native ds4

Motif-3 binder also accepted the completed mixed artifact as the official-final

53-layer, 14-full/39-SWA, 384E top-8 topology with MTP present.

Two final explicitly rebuilt sm_90 runtime repeats copied the full 94.16 GB

(87.70 GiB) image into one H200 in 9.560–12.070 seconds without SSD streaming

or CPU weight offload. The measured CUDA free-memory delta for model and runtime

initialization was 97,438,334,976–97,991,524,352 bytes; capacity accounting

uses the higher result. Strict residency fails startup instead of silently using

host-mapped weights.

Once optional CUDA preparation finishes, ds4 discards the raw GGUF tensor

pages while retaining only metadata/tokenizer mapping. Measured GGUF mapping

RSS fell from 91,955,608 kB to 9,416 kB and remained low through inference, so

the raw file is not kept as a second steady physical weight image beside the

CUDA-owned model copy.

The final all-sm_90 full-question 256K H200 attempt also measured 9,416 kB

of GGUF mapping RSS during partial prefill with VmSwap: 0, after allocating

the complete 262,144-token production latent cache.

The automated resident gate caps this mapping at 262,144 kB both after copy

and after native graph/cache execution. Its final native-sm_90 H200 run

measured 9,416 kB and 29,512–29,640 kB respectively.

The native expanded-path oracle and production latent path selected the same

first token and all top-8 logits on the short fixture; full-logit cosine was

0.99490164. Direct/chunked cache replay produced cosine 1.0. The real

mixed sparse-layer diagnostic measured Q2 down cosine 0.9996071 and final

sparse-output cosine 0.9998363. A real-weight MTP diagnostic evaluated 19

teacher-forced rows with finite logits.

DGX Spark ds4-dfm evidence

These measurements use one DGX Spark GB10 with 128 GB unified memory, NVIDIA

driver 610.43.02, CUDA 13.3, Linux 6.17, and the production 11-shard artifact

merged to its canonical 94,162,541,472-byte stream. Generation was greedy,

single-request, non-streaming, and used no speculative decoding or thinking.

The default prefill chunk was 4,096 tokens.

| Gate | Interface | Prefill | Decode | Correctness |

|---:|---|---:|---:|---|

| 8K | ds4-bench | 8,192 tokens at 519.55 tok/s | 64 tokens at 12.28 tok/s | prefill throughput point; prefill-only 519.55, decode-run 516.17 / 12.28 |

| 32K | OpenAI chat | 32,768 tokens in 82.649 s (396.47 tok/s) | 43 tokens in 4.799 s (8.96 tok/s) | exact beginning/middle/end JSON |

| 256K | OpenAI chat | 262,080 tokens in 1,492.375 s (175.61 tok/s) | 43 tokens in 17.072 s (2.52 tok/s) | exact beginning/middle/end JSON; finish_reason=stop |

The 256K request completed in 1,509.460 seconds with 262,123 total tokens.

Because the request was non-streaming, it does not provide an independently

measured network-visible TTFM.

The resident gate measured a 4,422,546,432-byte (4.119 GiB) latent-cache

payload and a 10,418,716,672-byte (9.703 GiB) default-graph-plus-cache physical

delta. Source GGUF RSS settled at 29,632 KiB; owner and worker both remained at

VmSwap: 0. The weight owner held 7.00 GiB of raw tensors plus 86.07 GiB in

590 derived artifacts, with no duplicate model image or CPU weight offload.

About 12 GiB remained available during the final gate.

Nsight Compute confirmed that the expanded attention kernel fell from 55.79

ms to 28.83 ms (48.3%), and the Motif group-5 QK-absorb kernel fell from 38.91

ms to 10.97 ms (71.8%). Loaded SM-clock samples were 2,398–2,411 MHz; the prior

611 MHz pinned-clock condition did not recur.

The table above contains single-request results. Concurrent 256K serving,

other GB10 software configurations, other accelerators, and model quality are

not claimed. The Motif ds4-dfm path uses plain decoding with --no-spec;

MTP and DSpark support models remain DeepSeek-only.

DGX Spark v0.6.2-dfm remesure

Same GB10 host (driver 610.43.02, CUDA 13.3, sm_121a), same MQ87-88

artifact, aligned-Q8 VMM owner (--reserve-gb 24), greedy, thinking

disabled. Engine:

Baekpica/ds4 dfm @ 6500480

(kernels through a09ff4f).

The integration tag remains v0.6.2-dfm;

that tag is not moved. The historical table above stays the

v0.5.6.3-dfm evidence.

| Gate | Interface | Prefill | Decode | Correctness |

|---|---|---:|---:|---|

| 8K | ds4-bench | 8,192 tokens at 627.19 tok/s | 64 tokens at 15.06 tok/s | throughput fixture |

| 32K | ds4-bench | 32,743 tokens at 545.62 tok/s | 64 tokens at 12.95 tok/s | same-session prior binary 519.57 / 13.00 |

| 32K | OpenAI chat | 32,751 tokens at 546.7 tok/s | 52 tokens at 12.8 tok/s | beginning/middle/end sentinels exact |

| 256K | OpenAI chat, -c 262144 | 262,080 tokens in 1,098.433 s (238.59 tok/s) | 43 tokens in 7.205 s (5.97 tok/s) | exact beginning/middle/end JSON; finish_reason=stop; 262,123 total; cached_tokens=0 |

Versus the merge-day v0.6.2-dfm baseline on this host (519.90 / 12.62 at

8K, 445.03 / 9.68 at 32K): 8K prefill +21%, 8K decode +19%, 32K

prefill +23%, 32K decode +34%. Versus the v0.5.6.3-dfm published

256K row (175.61 / 2.52): prefill +35.9%, decode +137%. Concurrent

256K banks are not claimed. The 256K worker held 10,429 MiB with 4.119 GiB

of latent KV; owner and worker VmSwap stayed 0; available memory stayed

11–12 GiB; SM clocks sampled 2,411–2,496 MHz.

Partial prefix reuse

Runtime

Baekpica/ds4@cf817c7

extends the shared partial-prefix checkpoint pool (introduced for Solar in

042bcea)

to the Motif-3 continuous lane. Motif-3 has no recurrent state, so a

checkpoint stores only each sliding-window layer's 128-row window (39

layers, 5.48 MiB per slot, 32 demand-mapped slots); full-attention latent

rows below the checkpoint copy directly from the retained source bank.

Request boundaries are checkpoints, long prefills and decode retain roughly

24 stride-aligned ones across the configured context, and a divergent

prompt restores the nearest checkpoint at or below the shared prefix and

replays only the gap.

One controlled A/B run used the same GB10, this artifact, the VMM owner

(--reserve-gb 16), a four-bank worker, -c 65536, 4,096-token prefill

chunks, and greedy no-think streaming Chat requests. The control set

DS4_SERVER_FORK_PARTIAL=0, which also skips checkpoint reservation; the

treatment used the default.

| Request | Prompt | Control cached | Treatment cached | Control TTFT | Treatment TTFT | Speedup |

|---|---:|---:|---:|---:|---:|---:|

| Source | 16,837 | 0 | 0 | 28,646.2 ms | 28,711.9 ms | 0.998× |

| Branch at ~7K | 7,140 | 0 | 4,096 | 11,319.4 ms | 5,196.3 ms | 2.18× |

| Branch at ~14K | 14,095 | 0 | 12,288 | 23,563.1 ms | 3,627.9 ms | 6.50× |

All three responses were byte-identical between control and treatment.

Treatment reported two partial-fork admits, 16,384 cached tokens, zero

failed requests, and zero admission rejects. Source checkpoint capture

added 65.7 ms (0.23%) in this single run. The engine fixture

(test_motif3_batch --partial-only) additionally pinned request-boundary

and periodic checkpoints and a two-segment ring-wrapping window restore,

each matching its cold oracle's greedy token.

This is a bounded worker-local cache, not an unbounded radix tree. Token

history and full-attention rows remain anchored to a retained bank, the

pool is not serialized into disk-KV payloads, and the EXAONE bank lane

keeps exact-frontier reuse only. Details:

docs/motif3-partial-reuse-2026-08-22.md.

DGX Spark 196K persistent multi-bank evidence

The current runtime was launched with a 196,608-token context, three resident

banks, an 8,192-token prefill chunk, and no speculative decoding. It completed

all four common endpoints--OpenAI Chat Completions, OpenAI Completions, OpenAI

Responses, and Anthropic Messages--with their native HTTP 200 response shapes.

| Gate | Verified result |

|---|---|

| 8K cold prefill | 8,214 prompt tokens at 266.3 tok/s; exact LONG_OK response |

| Single decode | 192 output tokens; 490.4 ms TTFT, 12.9 tok/s, 15.350 s HTTP wall time |

| Three simultaneous Chat requests | 192 output tokens each in 24.885--25.030 s; 23.01 aggregate output tok/s; served=3 fallback=0 |

Settled /v1/stats showed zero failed or serial requests, zero continuous-batch

failures, three total and zero live banks, and zero speculative drafts. The VMM

weight owner used 90,119 MiB, the worker used 22,283 MiB after the 8K checks,

and about 6.5 GiB of system memory remained available without an OOM event.

Loaded SM clock remained 2,411 MHz instead of the earlier faulty 611 MHz pin.

H200 end-to-end evidence

| Gate | Interface | Prefill | Decode | Correctness |

|---:|---|---:|---:|---|

| 2K | OpenAI chat | 346.72 tok/s | 12.64 tok/s | exact beginning/middle/end JSON |

| 32K | native, all-sm_90 | 125.34 tok/s | 1.942 tok/s | exact beginning/middle/end JSON; 43-token decode |

| 32K | OpenAI chat, all-sm_90 | 125.22 tok/s | 1.941 tok/s | exact JSON; model ID and 32,768 prompt tokens exact |

| 64K | native, all-sm_90 | 68.72 tok/s | 1.021 tok/s | exact beginning/middle/end JSON; 52-token decode |

| 128K | native, all-sm_90 | 36.36 tok/s | 0.524 tok/s | exact JSON; 131,072-token prompt + 49-token decode |

| 256K | native, legacy trim | 245,760/262,080 partial; 20.02 cumulative tok/s | not attempted | stopped for Spark handoff; not a correctness pass |

| 256K | native, all-sm_90, full question | 106,496/262,080 partial; 44.26 cumulative tok/s | not attempted | stopped for Spark handoff; not a correctness pass |

> [!CAUTION]

> These H200 figures are historical correctness-first bring-up measurements.

> The optimized Spark results above are separate and do not retroactively

> establish an H200 256K pass.

The isolated legacy-trim 256K row predates the final explicit sm_90 rebuild

and executes through the CUDA toolkit-compatible default code object, with MMQ

already at sm_90. Its rate is partial bring-up data, not a native-sm_90

performance claim. The final overlay has every CUDA code object verified as

sm_90, passes the full resident graph/cache regression, and passes the

native-sm_90 32K API plus native 32K/64K/128K rows shown above. The

full-question attempt used that all-sm_90 binary. At the user's direction,

both 256K attempts were stopped before decode and the remaining execution and

optimization were transferred to Spark. Precision, topology, and context were

not reduced to improve these figures.

The isolated 256K process also uses a legacy decode-reservation constant: it

keeps the final 20 tokens and omits the five-token prefix `QUESTION: Return

only a`, while retaining the full JSON/order instruction. Final ds4 revision

d878ea1 corrects the tail to 25 tokens. The handoff's separately hash-pinned

262,080-token OpenAI fixture removes only filler and preserves the complete

question; that corrected server gate remains mandatory on GB10.

The OpenAI server also completed a structured get_weather tool-call/result

loop. Its no-thinking continuation reused the full 165-token live prefix and

evaluated only the 51-token tool-result/new-assistant suffix. Two simultaneous

deterministic requests on two resident sessions both returned the expected

output, exercising continuous batching and session isolation.

Machine-readable structural and numerical reports are included as

MQ87-88-FIT-VERIFY.json and MQ87-88-FIT-SAMPLE-VERIFY.json.

The human-readable artifact and host records are included as

MIXED-QUANT.md and H200-DEVELOPMENT.md.

H200 development has validated the pinned source inventory, official router,

PolyNorm, mHC, tokenizer/chat/tool handling, expanded and latent GDLA paths,

the full Q8_0 GGUF, the 302,080-token Q8 activation-collection pass, strict

residency, short/long native generation, and the OpenAI server path above.

This card claims the completed strict single-DGX-Spark 262,080-token-prompt

plus decode gate and the separate three-bank 196K-context serving gate only

under the exact conditions above. It does not claim concurrent 256K serving,

equivalent performance on other hardware or software configurations, or

downstream task quality.

Limitations

• This is a hardware-oriented, very-low-bit routed-expert quantization. Quality

must be evaluated for the intended languages, reasoning, code, tools, and

long-context workloads.

• The public GGUF is a weight artifact, not a guarantee that an unrelated

runtime implements Motif-3 correctly.

• The 262,144-token metadata comes from the source architecture; the measured

validation in this card applies only to the pinned ds4-dfm configuration.

• DGX Spark uses coherent unified memory; conventional host-RAM plus discrete

VRAM accounting describes a different deployment.

Acknowledgements

• Motif Technologies — the

Motif-3 model and official implementation.

• antirez/ds4 — the original engine,

loader, server, session machinery, and routed-MoE foundation.

• Entrpi/ds4-on-spark — DGX Spark

CUDA and unified-memory groundwork used by downstream ds4 development.

• ggml-org/llama.cpp — GGUF and the

quantization formats used by this artifact.

Errors in this conversion, runtime port, calibration, or measurements are

ours, not theirs.

License and attribution

The source model identifies its license as MIT. See the official

Motif-Technologies/Motif-3 model card

for intended use, evaluation, citation, and license context.