Model benchmarks
flint-Q4_K_M local LLM performance
As of November 2026, flint-Q4_K_M runs at up to 37.0 tok/s for local inference (best of 2 community benchmark runs across 1 GPU).
Model size
40B
Peak speed
37.0 tok/s
Average speed
29.8 tok/s
Avg PP
274.4 tok/s
Min memory
25.1 GB
Max context
65,536 tokens
Avg output / run
25,505 tokens
Avg runtime / run
16m 1s
Avg quality
76.4
Benchmark runs
2
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 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 | 74.4 | 76.4 | 78.3 |
| Agent Workflow | 76.6 | 80.7 | 84.8 |
| Code Generation | 73.2 | 74.5 | 75.9 |
| Role Play & Narrative | 71.1 | 78.9 | 86.7 |
| Research & Analysis | 68.6 | 71.4 | 74.1 |
Performance by hardware and tool
Every hardware/tool/quantization combination flint-Q4_K_M 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 |
|---|---|---|---|---|---|---|---|---|
| Intel Arc Pro B390 | llama.cpp | Q4_K_M | 37.0 tok/s | 29.8 tok/s | 25.1 GB | 65,536 tokens | 76.4 | 2 |
Benchmark runs
All 2 flint-Q4_K_M runs submitted so far — expand one for its hardware, quality breakdown and per-scenario detail.
Frequently asked questions
- Is flint-Q4_K_M good for coding?
- In our benchmarks, flint-Q4_K_M scores 74.5/100 for coding. It runs at about 29.8 tok/s, so if you want more speed, local-qwen is faster (~60.2 tok/s) and still scores well for coding (84.8/100).
- Is flint-Q4_K_M good for agentic (tool-using) tasks?
- In our benchmarks, flint-Q4_K_M scores 80.7/100 for agentic workflows. It runs at about 29.8 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 flint-Q4_K_M for local inference?
- Across 2 community benchmark runs, flint-Q4_K_M reaches up to 37.0 tok/s and averages 29.8 tok/s, with the fastest results on Intel Arc Pro B390.
- How much memory does flint-Q4_K_M need?
- The leanest observed configuration used about 25.1 GB of memory (quantizations tested: Q4_K_M).
- Which tools have been used to run flint-Q4_K_M?
- Benchmarks were submitted using llama.cpp. Results are community-contributed and updated as new runs arrive.