1× DGX Spark
52.6tok/sEveryday
1 request, prose prompt, with DFlash2 (per recipe) Source
Intelligenceno independent value
- Custom kernel required
- Experimental
Local AI · Recipe · 1× DGX Spark
Ornith-1.5-122B EXL3 2.58 bpw with vLLM on 1× DGX Spark
Ornith-1.5-122B EXL3 2.58 bpw with vLLM on 1× DGX Spark: 30.5 tok/s according to github.com (dataset as of Nov 3, 2026).
by vcruz305
30.5tok/sPeak
Best value reported by source, 1 request, prompt type not stated, greedy, 512 tokens, default serve.sh without MTP
Source: github.comIntelligence (original model) · Artificial Analysisno independent value
≤ 3 bit: quantization may cost quality
Engine
- Engine
- vLLM 0.32.0 + vllm-exl3 Fork 3275819 (-DP2B_CB=2), VLLM_EXL3_COOP=1 + COALESCE_SHARDS=1, BF16-SSM
- Quantization
- EXL3 SAGE 2.58bpw (MixedK, mul1-Codebook, lm_head K=8), BF16 KV
- Model family
- Ornith-1.5-122B
- Context
- 262,144
- Parameters
- 122B MoE, 256 Experten top-8 + 1 Shared (laut README)
- Creator
- vcruz305
- GitHub stars
- 0
- Repo updated
- Oct 22, 2026
Measurements
Every sourced value of this recipe, each with its condition and source. Bars relative to the largest value in the group.
Decode · 1 request
Prefill by context
Time to first token
Shorter is better.
What you need
- Hardware
- 1 × NVIDIA DGX Spark (GB10)
- Weights
- Main weightsvcruz305/Ornith-1.5-122B-EXL3-SAGE-2.58bpw MIT
- Baseornith-ai/Ornith-1.5-122B MIT
- Engine
- vLLM 0.32.0 + vllm-exl3 Fork 3275819 (-DP2B_CB=2), VLLM_EXL3_COOP=1 + COALESCE_SHARDS=1, BF16-SSM
- Context
- 262,144 tokens
Notes
What matters before you rebuild it.
- ≤ 3 bit: quantization may cost qualityAt 3 bit and below the model may answer noticeably worse than the original. The intelligence number refers to the original.
- Custom kernel required
- tool-calling
Sources
- github.com/vcruz305/Ornith-1.5-122B-EXL3-DGX-Spark-recipe#measured-performance-on-1-nvidia-dgx-spark-gb10-sm_121-128-gb-unified
- github.com/vcruz305/Ornith-1.5-122B-EXL3-DGX-Spark-recipe#model
- github.com/vcruz305/Ornith-1.5-122B-EXL3-DGX-Spark-recipe/blob/main/HISTORICAL_BENCHMARKS.md
- huggingface.co/vcruz305/Ornith-1.5-122B-EXL3-SAGE-2.58bpw
Related recipes
1× DGX Spark
71.8tok/sEveryday
1 request, prose prompt (sampled, thinking off), with LiLiCorr (per recipe) Source
Intelligenceno independent value
- ≤ 3 bit: quantization may cost quality
- Custom kernel required
- License unclear
1× DGX Spark
60.3tok/sEveryday
1 request, prose prompt (essay), context 2k, no drafter (per recipe) Source
Intelligenceno independent value
- Custom kernel required
1× DGX Spark
59.4tok/sEveryday
1 request, chat prompt (20 prompts, thinking off), no drafting (default profile) Source
Intelligenceno independent value
- Custom kernel required
1× DGX Spark
58.3tok/sEveryday
1 request, prose prompt, with DFlash2 (per recipe) Source
Intelligence34with thinkingno thinking 20
- Custom kernel required
1× DGX Spark
55.6tok/sEveryday
1 request, prose prompt, with MTP (per recipe) Source
Intelligence40with thinking
- Custom kernel required
- License unclear
Local AI in your company?
In a workshop we work out which models and which hardware fit your tasks, and build the first agent on your infrastructure.