Model benchmarks
DavidAU/Llama-3.2-8X3B-MOE-Dark-Champion-Instruct-uncensored-abliterated-18.4B-GGUF local LLM performance
As of October 2026, DavidAU/Llama-3.2-8X3B-MOE-Dark-Champion-Instruct-uncensored-abliterated-18.4B-GGUF runs at up to 8.9 tok/s for local inference (best of 5 community benchmark runs across 1 GPU).
Model size
18.4B
Peak speed
8.9 tok/s
Average speed
5.2 tok/s
Avg PP
147.5 tok/s
Min memory
n/a
Max context
131,070 tokens
Avg output / run
4,199 tokens
Avg runtime / run
15m 32s
Avg quality
34.3
Benchmark runs
5
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 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 | 27.3 | 34.3 | 40.3 |
| Agent Workflow | 14.0 | 30.5 | 45.2 |
| Code Generation | 4.5 | 14.3 | 21.4 |
| Role Play & Narrative | 45.2 | 55.8 | 66.2 |
| Research & Analysis | 33.9 | 36.8 | 39.9 |
Performance by hardware and tool
Every hardware/tool/quantization combination DavidAU/Llama-3.2-8X3B-MOE-Dark-Champion-Instruct-uncensored-abliterated-18.4B-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 |
|---|---|---|---|---|---|---|---|---|
| Intel Arc B390 | llama.cpp | — | 8.9 tok/s | 5.2 tok/s | n/a | 131,070 tokens | 34.3 | 5 |
Benchmark runs
All 5 DavidAU/Llama-3.2-8X3B-MOE-Dark-Champion-Instruct-uncensored-abliterated-18.4B-GGUF runs submitted so far — expand one for its hardware, quality breakdown and per-scenario detail.
Frequently asked questions
- Is DavidAU/Llama-3.2-8X3B-MOE-Dark-Champion-Instruct-uncensored-abliterated-18.4B-GGUF good for coding?
- In our benchmarks, DavidAU/Llama-3.2-8X3B-MOE-Dark-Champion-Instruct-uncensored-abliterated-18.4B-GGUF scores 14.3/100 for coding. It runs at about 5.2 tok/s, so if you want more speed, IsValorum/Nex-N2.5-mini-APEX-I-MiniPlus-V2.1-GGUF:1 is faster (~25.7 tok/s) and still scores well for coding (85.0/100).
- Is DavidAU/Llama-3.2-8X3B-MOE-Dark-Champion-Instruct-uncensored-abliterated-18.4B-GGUF good for agentic (tool-using) tasks?
- In our benchmarks, DavidAU/Llama-3.2-8X3B-MOE-Dark-Champion-Instruct-uncensored-abliterated-18.4B-GGUF scores 30.5/100 for agentic workflows. It runs at about 5.2 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 DavidAU/Llama-3.2-8X3B-MOE-Dark-Champion-Instruct-uncensored-abliterated-18.4B-GGUF for local inference?
- Across 5 community benchmark runs, DavidAU/Llama-3.2-8X3B-MOE-Dark-Champion-Instruct-uncensored-abliterated-18.4B-GGUF reaches up to 8.9 tok/s and averages 5.2 tok/s, with the fastest results on Intel Arc B390.
- Which tools have been used to run DavidAU/Llama-3.2-8X3B-MOE-Dark-Champion-Instruct-uncensored-abliterated-18.4B-GGUF?
- Benchmarks were submitted using llama.cpp. Results are community-contributed and updated as new runs arrive.