Model benchmarks
protoLabsAI/Ornith-1.5-9B-MTP-GGUF local LLM performance
As of October 2026, protoLabsAI/Ornith-1.5-9B-MTP-GGUF runs at up to 92.5 tok/s for local inference (best of 5 community benchmark runs across 2 GPUs).
Model size
9B
Peak speed
92.5 tok/s
Average speed
60.5 tok/s
Avg PP
940.3 tok/s
Min memory
10.6 GB
Max context
65,536 tokens
Avg output / run
49,697 tokens
Avg runtime / run
16m 6s
Avg quality
70.9
Benchmark runs
5
GPUs tested
2
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 5 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 | 59.3 | 70.9 | 77.9 |
| Agent Workflow | 61.4 | 75.2 | 88.9 |
| Code Generation | 1.9 | 44.0 | 78.8 |
| Role Play & Narrative | 70.4 | 80.7 | 88.3 |
| Research & Analysis | 79.2 | 83.6 | 87.4 |
Performance by hardware and tool
Every hardware/tool/quantization combination protoLabsAI/Ornith-1.5-9B-MTP-GGUF 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 |
|---|---|---|---|---|---|---|---|---|
| NVIDIA GeForce RTX 4070 | Unsloth Studio | IQ4_XS | 92.5 tok/s | 92.5 tok/s | 10.6 GB | 65,536 tokens | 78.1 | 1 |
| NVIDIA GeForce RTX 4070 | Unsloth Studio | Q4_K_M | 88.7 tok/s | 88.7 tok/s | n/a | 65,536 tokens | 56.8 | 1 |
| NVIDIA GeForce RTX 4070 | Unsloth Studio | Q5_K_M | 83.7 tok/s | 83.7 tok/s | n/a | 52,000 tokens | 69.1 | 1 |
| Intel(R) Arc(TM) 140V GPU (16GB) | Unsloth Studio | IQ4_XS | 19.7 tok/s | 19.7 tok/s | 15.4 GB | 16,384 tokens | 73.4 | 1 |
| Intel(R) Arc(TM) 140V GPU (16GB) | Unsloth Studio | Q4_K_M | 18.0 tok/s | 18.0 tok/s | 18.3 GB | 65,536 tokens | 77.0 | 1 |
Benchmark runs
All 5 protoLabsAI/Ornith-1.5-9B-MTP-GGUF runs submitted so far — expand one for its hardware, quality breakdown and per-scenario detail.
Frequently asked questions
- Is protoLabsAI/Ornith-1.5-9B-MTP-GGUF good for coding?
- In our benchmarks, protoLabsAI/Ornith-1.5-9B-MTP-GGUF scores 44.0/100 for coding. It runs at about 60.5 tok/s, so if you want more speed, ornith-1.5-35b-a3b-apex-i-quality is faster (~73.2 tok/s) and still scores well for coding (83.1/100).
- Is protoLabsAI/Ornith-1.5-9B-MTP-GGUF good for agentic (tool-using) tasks?
- In our benchmarks, protoLabsAI/Ornith-1.5-9B-MTP-GGUF scores 75.2/100 for agentic workflows. It runs at about 60.5 tok/s, so if you want more speed, Nex-N2.5-mini-APEX-Mini is faster (~88.2 tok/s) and still scores well for agentic workflows (91.3/100).
- How fast is protoLabsAI/Ornith-1.5-9B-MTP-GGUF for local inference?
- Across 5 community benchmark runs, protoLabsAI/Ornith-1.5-9B-MTP-GGUF reaches up to 92.5 tok/s and averages 60.5 tok/s, with the fastest results on NVIDIA GeForce RTX 4070.
- How much memory does protoLabsAI/Ornith-1.5-9B-MTP-GGUF need?
- The leanest observed configuration used about 10.6 GB of memory (quantizations tested: IQ4_XS, Q4_K_M, Q5_K_M).
- Which tools have been used to run protoLabsAI/Ornith-1.5-9B-MTP-GGUF?
- Benchmarks were submitted using Unsloth Studio. Results are community-contributed and updated as new runs arrive.