Model benchmarks
ornith:35b local LLM performance
As of October 2026, ornith:35b runs at up to 56.4 tok/s for local inference (best of 3 community benchmark runs across 2 GPUs).
Model size
34.7B
Peak speed
56.4 tok/s
Average speed
41.4 tok/s
Avg PP
332.0 tok/s
Min memory
20.9 GB
Max context
180,000 tokens
Avg output / run
43,628 tokens
Avg runtime / run
22m 43s
Avg quality
71.8
Benchmark runs
3
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 3 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 | 61.7 | 71.8 | 80.8 |
| Agent Workflow | 18.6 | 55.4 | 80.4 |
| Code Generation | 65.2 | 72.7 | 79.1 |
| Role Play & Narrative | 75.0 | 81.8 | 88.4 |
| Research & Analysis | 69.1 | 77.2 | 83.2 |
Performance by hardware and tool
Every hardware/tool/quantization combination ornith:35b 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 7900 XTX | Ollama | Q4_K_M | 56.4 tok/s | 45.9 tok/s | 20.9 GB | 180,000 tokens | 66.9 | 2 |
| NVIDIA GeForce RTX 3070 | llama.cpp | — | 32.4 tok/s | 32.4 tok/s | n/a | 32,768 tokens | 81.6 | 1 |
Benchmark runs
All 3 ornith:35b runs submitted so far — expand one for its hardware, quality breakdown and per-scenario detail.
Frequently asked questions
- Is ornith:35b good for coding?
- In our benchmarks, ornith:35b scores 72.7/100 for coding. It runs at about 41.4 tok/s, so if you want more speed, K2-Horizon-MoVA-36B-A4B-MLX-4bit is faster (~48.5 tok/s) and still scores well for coding (82.5/100).
- Is ornith:35b good for agentic (tool-using) tasks?
- In our benchmarks, ornith:35b scores 55.4/100 for agentic workflows. It runs at about 41.4 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 ornith:35b for local inference?
- Across 3 community benchmark runs, ornith:35b reaches up to 56.4 tok/s and averages 41.4 tok/s, with the fastest results on AMD Radeon RX 7900 XTX.
- How much memory does ornith:35b need?
- The leanest observed configuration used about 20.9 GB of memory (quantizations tested: Q4_K_M).
- Which tools have been used to run ornith:35b?
- Benchmarks were submitted using Ollama, llama.cpp. Results are community-contributed and updated as new runs arrive.