Model benchmarks
unsloth/flint-GGUF:UD-Q4_K_XL local LLM performance
As of November 2026, unsloth/flint-GGUF:UD-Q4_K_XL runs at up to 200.9 tok/s for local inference (best of 2 community benchmark runs across 2 GPUs).
Model size
40B
Peak speed
200.9 tok/s
Average speed
166.2 tok/s
Avg PP
2657.2 tok/s
Min memory
25.3 GB
Max context
131,972 tokens
Avg output / run
31,476 tokens
Avg runtime / run
3m 30s
Avg quality
73.2
Benchmark runs
2
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 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.
| Task | P5 (low) | Avg | P95 (high) |
|---|---|---|---|
| Overall | 71.6 | 73.2 | 74.8 |
| Agent Workflow | 83.0 | 84.9 | 86.7 |
| Code Generation | 44.9 | 57.1 | 69.2 |
| Role Play & Narrative | 76.6 | 81.3 | 85.9 |
| Research & Analysis | 66.5 | 69.6 | 72.7 |
Performance by hardware and tool
Every hardware/tool/quantization combination unsloth/flint-GGUF:UD-Q4_K_XL has been benchmarked on, ranked by peak token generation speed. Last updated November 2026.
| Hardware | Tool | Quant | Best tok/s | Avg tok/s | Memory | Context | Quality | Runs |
|---|---|---|---|---|---|---|---|---|
| Apple M5 Max | splash | Q4_K | 200.9 tok/s | 200.9 tok/s | 25.3 GB | 131,972 tokens | 71.4 | 1 |
| NVIDIA GeForce RTX 3080 Ti | llama.cpp | Q4_K | 131.5 tok/s | 131.5 tok/s | 28.2 GB | 131,072 tokens | 75.0 | 1 |
Benchmark runs
All 2 unsloth/flint-GGUF:UD-Q4_K_XL runs submitted so far — expand one for its hardware, quality breakdown and per-scenario detail.
Frequently asked questions
- Is unsloth/flint-GGUF:UD-Q4_K_XL good for agentic (tool-using) tasks?
- In our benchmarks, unsloth/flint-GGUF:UD-Q4_K_XL scores 84.9/100 for agentic workflows. It runs at about 166.2 tok/s, so if you want more speed, swift-1.5-qwen3.8-flash-next-iq3_s is faster (~259.0 tok/s) and still scores well for agentic workflows (88.9/100).
- How fast is unsloth/flint-GGUF:UD-Q4_K_XL for local inference?
- Across 2 community benchmark runs, unsloth/flint-GGUF:UD-Q4_K_XL reaches up to 200.9 tok/s and averages 166.2 tok/s, with the fastest results on Apple M5 Max.
- How much memory does unsloth/flint-GGUF:UD-Q4_K_XL need?
- The leanest observed configuration used about 25.3 GB of memory (quantizations tested: Q4_K).
- Which tools have been used to run unsloth/flint-GGUF:UD-Q4_K_XL?
- Benchmarks were submitted using llama.cpp, splash. Results are community-contributed and updated as new runs arrive.