---

base_model: deepreinforce-ai/Ornith-1.0-35B

library_name: gguf

pipeline_tag: text-generation

license: other

license_name: mit-apache-2.0-composite

tags:

- gguf

- lm-studio

- llama.cpp

- ornith

- mixture-of-experts

- windows

- cuda

- nvidia

---

Ornith-1.0-35B Strix Halo Hybrid — LM Studio GGUF

> LM Studio compatible, no integrated MTP.

Choose the correct build

| Build | Use it with | Integrated MTP |

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

| Full optimized MTP build | Current llama.cpp with draft-mtp; patched Vulkan/CUDA deployments | Yes |

| LM Studio compatible build (this repository) | LM Studio runtimes that reject the integrated MTP block | No |

This repository is the compatibility build. It preserves all 40 optimized

base-model layers and removes only the appended MTP prediction layer and its

qwen35moe.nextn_predict_layers metadata. No remaining tensor was converted,

retrained, or requantized during the compatibility rewrite.

File

| File | Size | Quantization | SHA-256 |

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

| ornith-1.0-35b-hybrid-Q4_0-LMStudio.gguf | 18,866.19 MiB (18.424 GiB) | Q4_K_M with 20 hot ffn_down_exps tensors overridden to Q4_0; no MTP tensors | 02c057b9a7fc5abeae73b7c9b6774928a0f8fc01c44cb858dd715babdd2b851f |

Compared with the full MTP artifact, this file is 521.38 MiB smaller. The

removed data consists of exactly 20 blk.40.* prediction tensors. The output

contains 733 tensors, reports qwen35moe.block_count = 40, and contains no

blk.40.* or nextn_predict_layers entry.

Why a separate LM Studio build exists

The full artifact loads in newer llama.cpp builds that support integrated

Qwen3.6/Ornith MTP. LM Studio runtime 2.14.0 used for the Windows validation is

based on llama.cpp b8861 and rejected that appended block before allocating the

model. Removing only the unsupported MTP layer made the same optimized base

load successfully.

Validation on Windows 11:

• LM Studio runtime: llama.cpp-win-x86_64-nvidia-cuda12-avx2@2.14.0;
• llama.cpp revision reported by LM Studio: b8861 (cf8b0db);
• model load completed in 15.16 seconds in the compatibility smoke profile;
• a local generation request completed successfully;
• the incompatible full-MTP file failed in both GPU and CPU-only load tests,

proving that the failure was format/runtime support rather than VRAM.

Optimization retained in this build

• the compatible Ornith/Qwen3.6 35B A3B MoE architecture and chat template;
• selective Q4_0 replacement of 20 frequently streamed MoE down-projection

tensors, with the surrounding Q4_K_M quantization plan unchanged;

• all attention, dense, embedding, normalization and output tensors from the

optimized 40-layer base;

• the same inference quality as the full artifact when speculation is off.

No weights were trained or fine-tuned. Selective requantization changes the

numerical representation, so the original project checked quality end-to-end

on the same Orion coding slice.

Performance scope

The table below was measured on the same Windows 11 laptop with the official

llama.cpp b10066 CUDA container, 16,384 context, Q8_0 K/V cache, batch 2048,

ubatch 512, parallelism 1, dense/attention tensors on an RTX 4060 Laptop 8 GiB,

and MoE experts in host RAM. The Hybrid, speculation off row executes the

same 40-layer tensor set published here, but it is not presented as a direct

LM Studio-runtime benchmark.

| Laptop profile | 1K decode | 8K decode | Repeated-code decode | 8K prompt processing |

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

| Baseline Ornith Q4_K_M | 5.57 tok/s | 23.90 tok/s | 21.95 tok/s | 341.48 tok/s |

| This 40-layer hybrid tensor set, speculation off | 29.70 tok/s | 29.59 tok/s | 26.75 tok/s | 365.26 tok/s |

| Full linked build, MTP + n-gram | 28.74 tok/s | 66.88 tok/s | 67.23 tok/s | 354.19 tok/s |

The stable non-speculative improvement was +23.8% at 8K and +21.8% on

repeated code. The 60–67 tok/s result belongs to the linked MTP build under a

compatible runtime and must not be expected from this no-MTP file.

Direct LM Studio native API benchmark

The compatibility file was then measured through LM Studio's own

/api/v0/completions endpoint on the same Windows laptop. Each row generated

128 tokens at temperature 0. The 1K and 8K prompts were run twice; the repeated

8K code prompt was run three times. LM Studio reported decode throughput and

TTFT directly in the response. The first TTFT is cold-prompt processing; the

second shows LM Studio's prompt-cache path.

| LM Studio 2.14.0 profile | Actual prompt | Decode mean | Cold TTFT | Warm-cache TTFT |

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

| 1K prompt | 1,056 tokens | 33.48 tok/s | 4.296 s | 0.139 s |

| 8K prompt | 8,448 tokens | 34.77 tok/s | 23.160 s | 0.166 s |

| Repeated 8K code | 8,448 tokens | 33.25 tok/s | — | 0.151–0.170 s |

Against the b10066 no-speculation run on the same tensor set, LM Studio decode

was +12.7% at 1K, +17.5% at 8K and +24.3% on repeated code. These are measured

runtime differences, not a promise for other LM Studio versions or hardware.

| LM Studio telemetry | Peak |

|---|---:|

| GPU temperature | 66 C |

| GPU utilization | 91% |

| VRAM used | 2,329 MiB |

| GPU power | 45.1 W |

| Whole-system RAM used | 44.10 GiB |

The complete native response statistics are published in

lmstudio-native-benchmark.json.

Orion four-scenario coding slice through LM Studio

The same seeded workspaces, Orion binary, prompts and deterministic checks used

for the earlier laptop A/B were executed through

http://127.0.0.1:1234/v1. Settings remained approval=auto, 24 agent steps,

16,384 context and a 900-second outer timeout.

| Orion coding scenario | Score | Checks | Wall time |

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

| Java service coverage | 80% | 8/10 | 900.0 s (outer timeout) |

| Kafka Node order pipeline | 75% | 6/8 | 383.2 s |

| Rabbit retry / DLQ | 75% | 6/8 | 363.4 s |

| Kafka Java outbox / idempotency | 80% | 8/10 | 530.9 s |

| Total / average | 77.5% | 28/36 | 2,177.6 s (36m 17.6s) |

Quality is identical to both the baseline and full-hybrid laptop runs. Raw

decode was faster, but end-to-end agent time was longer because Orion repeated

verification-guard steps; the Java coverage scenario reached the 900-second

outer timeout after its file changes and tests were already complete.

The Rabbit post-run verifier initially failed to spawn the generic python

command with EPERM. The saved workspace compiled successfully and the exact

three-test suite passed under the bundled Python interpreter. The raw 5/8

result and original error are retained in the local research artifact; the

published summary records the recovered 6/8 score and that the model itself was

not rerun. orion-lmstudio-summary.json contains the check-level result.

Suggested LM Studio load profile

• context length: 16,384 for the first validated load;
• K cache: Q8_0;
• V cache: Q8_0;
• parallel sessions: 1;
• evaluation batch size: 2048;
• keep MoE expert layers in host RAM on an 8 GiB GPU;
• offload dense/attention layers to the GPU as memory permits.

Start at 4K or 8K context if LM Studio's resource guardrails reject the first

load, then increase context after measuring available RAM and VRAM. Context

length affects KV memory but does not restore MTP; use the linked full build

with a current llama.cpp runtime when speculative MTP acceleration is required.

Sources and licensing

The base model is

deepreinforce-ai/Ornith-1.0-35B.

The full linked build also contains compatible MTP tensors from

a4lg/Qwen3.6-35B-A3B-MTP-ONLY-GGUF.

This compatibility artifact contains no MTP tensors, but provenance is retained

because it was derived from that combined artifact. Review the source

repositories and the accompanying license files before redistribution or

commercial use.