Brooooooklyn/Ornith-1.0-35B-UD-Q5_K_XL-mlx

Ornith-1.0-35B — UD-Q5_K_XL (mlx-node)

5-bit base mixed-precision quantization of deepreinforce-ai/Ornith-1.0-35B for Apple Silicon, using the Unsloth Dynamic per-tensor bit allocation (without imatrix AWQ) via mlx-node.

Ornith-1.0 is a self-improving family of open-source agentic coding models. The 35B member is a Qwen3.5-VL-MoE (hybrid Gated-DeltaNet + full attention, 256 experts, vision-language) post-train.

	Original (BF16)	This Model
Size	~68 GB	26 GB
Format	SafeTensors (sharded)	SafeTensors (sharded)
Precision	BF16 uniform	Mixed 5/6/8/8-bit affine + BF16

All Variants

Repo	Format	Size	Decode (tok/s)
Brooooooklyn/Ornith-1.0-35B-UD-Q3_K_XL-mlx	UD-Q3_K_XL	17 GB	111.6
Brooooooklyn/Ornith-1.0-35B-mxfp4-mlx	MXFP4	20 GB	107.8
Brooooooklyn/Ornith-1.0-35B-UD-Q4_K_XL-mlx	UD-Q4_K_XL	22 GB	102.3
Brooooooklyn/Ornith-1.0-35B-nvfp4-mlx	NVFP4	23 GB	94.6
Brooooooklyn/Ornith-1.0-35B-UD-Q5_K_XL-mlx (this model)	UD-Q5_K_XL	26 GB	95.4
Brooooooklyn/Ornith-1.0-35B-UD-Q6_K_XL-mlx	UD-Q6_K_XL	31 GB	93.1
Brooooooklyn/Ornith-1.0-35B-UD-Q8_K_XL-mlx	UD-Q8_K_XL	36 GB	91.5
Brooooooklyn/Ornith-1.0-35B-mxfp8-mlx	MXFP8	36 GB	84.8

Benchmarked on a cool Apple M5 Max: median decode throughput over three 512-token generations, with a 60-second idle GPU cooldown after every generation. (Sustained decode on Apple Silicon is thermally sensitive — back-to-back benchmarking on a hot chip can understate throughput by 20–30%, so every model here was measured from a comparable cool start.)

Performance

Steady-state decode: 95.4 tok/s (1.5x vs BF16) on Apple M5 Max. Decode is memory-bandwidth bound on Apple Silicon — fewer bytes per token directly translates to higher throughput. The MoE architecture activates only 8 of 256 experts per token (~3B active out of 35.9B total), so the active-weight footprint streamed per token is what matters.

Output Quality

Decoded-text quality was verified against the BF16 reference with a multi-judge review of the actual generated output (not a heuristic): a 4-turn factual chat plus a Python is_balanced() bracket-matching task. This UD-Q5_K_XL build produced coherent prose, correct facts, and a correct implementation — no runaway generation, repetition loops, or stray tokens — on par with full precision. (The 2-bit tier is intentionally excluded from this collection: it was the only width that showed coherence breakdown.)

Per-Tensor Quantization

Weight	Bits	Rationale
embed_tokens	8-bit affine	KLD ~0.15 — very low sensitivity
lm_head	8-bit affine	KLD ~0.05 — safest tensor
self_attn.q/k/v_proj	8-bit affine	KLD ~1.5–2.9 — attention-sensitive
linear_attn.in_proj_qkv/z	8-bit affine	KLD ~2.9 — SSM input gates
self_attn.o_proj	8-bit affine	KLD ~1.5; row-independent qmv for T=0 exactness
linear_attn.out_proj	8-bit affine	KLD ~6.0 — worst tensor; kept high
linear_attn.in_proj_a/b	8-bit affine	tiny low-rank GDN projections
switch_mlp.down_proj	6-bit affine	"slightly more sensitive" than other FFN
switch_mlp.gate_proj/up_proj	5-bit affine	bulk of the expert budget
Router gates (mlp.gate, shared_expert_gate)	8-bit affine	MoE routing accuracy
GDN params (A_log, dt_bias)	bf16	state-space dynamics
vision_tower.*	bf16	vision encoder kept full precision

Quantization Strategy

Built on Unsloth Dynamic 2.0 per-tensor KLD analysis: sensitive layers (attention/SSM inputs, down_proj, embeddings/head) get higher bits, while the bulk of FFN expert weights are quantized to the base width. self_attn.o_proj, linear_attn.out_proj, the split low-rank GDN projections (in_proj_a/b) and the MoE router gates are pinned to 8-bit affine (group_size 64). GatedDeltaNet state-space parameters and the vision encoder stay bf16.

Note: These ornith quants apply the Unsloth bit allocation without imatrix AWQ pre-scaling — ornith has no published imatrix, so the attention/SSM channels are quantized directly. Expect a small quality gap versus an imatrix-calibrated build at the lowest bit widths.

Architecture

Parameter	Value
Total parameters	35.9B (~3B active per token)
Hidden size	2,048
Layers	40 (30 linear GatedDeltaNet + 10 full attention)
Attention heads	16 (2 KV heads, GQA 8:1)
Head dimension	256
Experts	256 per MoE layer, top-8 routing
Vocab size	248,320
Vision	yes (Qwen3.5-VL vision tower, kept bf16)
Max context	262,144 tokens

Usage

import { loadSession } from '@mlx-node/lm';

const session = await loadSession('./Ornith-1.0-35B-UD-Q5_K_XL-mlx');

for await (const event of session.sendStream('Write a Python function to merge two sorted lists.', {
  config: { maxNewTokens: 2048, temperature: 0.6, reasoningEffort: 'low' },
})) {
  if (!event.done) process.stdout.write(event.text);
}

How It Was Made

mlx convert \
  -i Ornith-1.0-35B \
  -o Ornith-1.0-35B-UD-Q5_K_XL-mlx \
  -q --q-recipe unsloth --q-bits 5

The Unsloth recipe's per-tensor bit tiers were applied without imatrix AWQ (no native ornith imatrix). 7-bit tiers are snapped up to 8-bit (MLX affine supports 2/3/4/5/6/8-bit).

Acknowledgments

• Unsloth — Per-layer KLD bit-allocation strategy (Dynamic 2.0)
• DeepReinforce — For the Ornith-1.0 model family
• Qwen Team — For the Qwen3.5 base architecture
• Apple MLX — For the Metal-accelerated ML framework

License

MIT (inherited from base model).