Model benchmarks
pfeifferj/Ornith-1.5-35B-A3B-GSQ-RCO-GGUF:3bit local LLM performance
As of October 2026, pfeifferj/Ornith-1.5-35B-A3B-GSQ-RCO-GGUF:3bit runs at up to 103.8 tok/s for local inference (best of 10 community benchmark runs across 1 GPU).
Model size
35B
Peak speed
103.8 tok/s
Average speed
88.5 tok/s
Avg PP
1186.6 tok/s
Min memory
12.3 GB
Max context
131,072 tokens
Avg output / run
29,115 tokens
Avg runtime / run
5m 46s
Avg quality
77.4
Benchmark runs
10
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 10 runs. The low and high columns show how much the judge’s score varies between runs, and need at least two runs to display.
| Task | P5 (low) | Avg | P95 (high) |
|---|---|---|---|
| Overall | 63.6 | 77.4 | 85.7 |
| Agent Workflow | 12.7 | 63.7 | 90.1 |
| Code Generation | 63.6 | 73.9 | 81.8 |
| Role Play & Narrative | 79.4 | 86.9 | 93.3 |
| Research & Analysis | 82.6 | 85.1 | 88.1 |
Performance by hardware and tool
Every hardware/tool/quantization combination pfeifferj/Ornith-1.5-35B-A3B-GSQ-RCO-GGUF:3bit has been benchmarked on, ranked by peak token generation speed. Last updated October 2026.
| Hardware | Tool | Quant | Best tok/s | Avg tok/s | Memory | Context | Quality | Runs |
|---|---|---|---|---|---|---|---|---|
| AMD Radeon RX 9070/9070 XT/9070 GRE | llama.cpp | — | 103.8 tok/s | 88.4 tok/s | 13.3 GB | 32,768 tokens | 77.1 | 6 |
| AMD Radeon RX 9070/9070 XT/9070 GRE | Ollama | Q3_K_S | 102.5 tok/s | 88.6 tok/s | 12.3 GB | 131,072 tokens | 77.9 | 4 |
Benchmark runs
All 10 pfeifferj/Ornith-1.5-35B-A3B-GSQ-RCO-GGUF:3bit runs submitted so far — expand one for its hardware, quality breakdown and per-scenario detail.
Frequently asked questions
- Is pfeifferj/Ornith-1.5-35B-A3B-GSQ-RCO-GGUF:3bit good for coding?
- In our benchmarks, pfeifferj/Ornith-1.5-35B-A3B-GSQ-RCO-GGUF:3bit scores 73.9/100 for coding. It runs at about 88.5 tok/s, so if you want more speed, ornith-1.5-35b-a3b-apex-mtp-i-quality is faster (~105.2 tok/s) and still scores well for coding (83.6/100).
- Is pfeifferj/Ornith-1.5-35B-A3B-GSQ-RCO-GGUF:3bit good for agentic (tool-using) tasks?
- In our benchmarks, pfeifferj/Ornith-1.5-35B-A3B-GSQ-RCO-GGUF:3bit scores 63.7/100 for agentic workflows. It runs at about 88.5 tok/s, so if you want more speed, ornith-1.5-35b-a3b-apex-mtp-quality is faster (~103.2 tok/s) and still scores well for agentic workflows (88.5/100).
- How fast is pfeifferj/Ornith-1.5-35B-A3B-GSQ-RCO-GGUF:3bit for local inference?
- Across 10 community benchmark runs, pfeifferj/Ornith-1.5-35B-A3B-GSQ-RCO-GGUF:3bit reaches up to 103.8 tok/s and averages 88.5 tok/s, with the fastest results on AMD Radeon RX 9070/9070 XT/9070 GRE.
- How much memory does pfeifferj/Ornith-1.5-35B-A3B-GSQ-RCO-GGUF:3bit need?
- The leanest observed configuration used about 12.3 GB of memory (quantizations tested: Q3_K_S).
- Which tools have been used to run pfeifferj/Ornith-1.5-35B-A3B-GSQ-RCO-GGUF:3bit?
- Benchmarks were submitted using Ollama, llama.cpp. Results are community-contributed and updated as new runs arrive.