Model benchmarks

Ornith-1.5-35B-A3B-APEX-MTP-I-Compact local LLM performance

As of September 2026, Ornith-1.5-35B-A3B-APEX-MTP-I-Compact runs at up to 40.0 tok/s for local inference (best of 2 community benchmark runs across 1 GPU).

LM StudiovLLM
ShareRedditX

Model size

35B

Peak speed

40.0 tok/s

Average speed

31.5 tok/s

Avg PP

603.5 tok/s

Min memory

n/a

Max context

131.072 tokens

Avg output / run

26.586 tokens

Avg runtime / run

13m 57s

Avg quality

83.7

Benchmark runs

2

GPUs tested

1

Quality by task

Average LLM-judged quality (0–100) with the run-to-run spread shown as a P5–P95 band, overall and for each benchmark task, across all 2 runs. The low and high columns show how much the judge’s score varies between runs, and need at least two runs to display.

TaskP5 (low)AvgP95 (high)
Overall81.983.785.4
Agent Workflow85.588.190.7
Code Generation79.479.679.8
Role Play & Narrative79.984.689.3
Research & Analysis82.382.482.4

Performance by hardware and tool

Every hardware/tool/quantization combination Ornith-1.5-35B-A3B-APEX-MTP-I-Compact has been benchmarked on, ranked by peak token generation speed. Last updated September 2026.

HardwareToolQuantBest tok/sAvg tok/sMemoryContextQualityRuns
AMD Radeon RX 9050 / 9060 XTLM Studio—40.0 tok/s40.0 tok/s16.2 GB131.072 tokens85.61
CPU onlyvLLM—23.0 tok/s23.0 tok/sn/a98.304 tokens81.71

Benchmark runs

All 2 Ornith-1.5-35B-A3B-APEX-MTP-I-Compact runs submitted so far — expand one for its hardware, quality breakdown and per-scenario detail.

Frequently asked questions

Is Ornith-1.5-35B-A3B-APEX-MTP-I-Compact good for coding?
In our benchmarks, Ornith-1.5-35B-A3B-APEX-MTP-I-Compact scores 79.6/100 for coding. It runs at about 31.5 tok/s, so if you want more speed, unsloth/Qwen3.8-27B-GGUF:IQ3_S is faster (~104.3 tok/s) and still scores well for coding (83.0/100).
Is Ornith-1.5-35B-A3B-APEX-MTP-I-Compact good for agentic (tool-using) tasks?
In our benchmarks, Ornith-1.5-35B-A3B-APEX-MTP-I-Compact scores 88.1/100 for agentic workflows. It runs at about 31.5 tok/s, so if you want more speed, granite-4.2-8b-Q4_K_L is faster (~71.6 tok/s) and still scores well for agentic workflows (92.9/100).
How fast is Ornith-1.5-35B-A3B-APEX-MTP-I-Compact for local inference?
Across 2 community benchmark runs, Ornith-1.5-35B-A3B-APEX-MTP-I-Compact reaches up to 40.0 tok/s and averages 31.5 tok/s, with the fastest results on AMD Radeon RX 9050 / 9060 XT.
How much memory does Ornith-1.5-35B-A3B-APEX-MTP-I-Compact need?
The leanest observed configuration used about n/a of memory.
Which tools have been used to run Ornith-1.5-35B-A3B-APEX-MTP-I-Compact?
Benchmarks were submitted using LM Studio, vLLM. Results are community-contributed and updated as new runs arrive.