Model benchmarks
Flint_t1p95k20m0 local LLM performance
As of November 2026, Flint_t1p95k20m0 runs at up to 90.7 tok/s for local inference (best of 2 community benchmark runs across 1 GPU).
Model size
2.1B
Peak speed
90.7 tok/s
Average speed
88.9 tok/s
Avg PP
581.5 tok/s
Min memory
58.0 GB
Max context
32,768 tokens
Avg output / run
24,634 tokens
Avg runtime / run
4m 33s
Avg quality
75.5
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 | 72.6 | 75.5 | 78.5 |
| Agent Workflow | 76.4 | 78.6 | 80.7 |
| Code Generation | 68.0 | 70.0 | 72.0 |
| Role Play & Narrative | 61.6 | 75.8 | 90.0 |
| Research & Analysis | 76.0 | 77.7 | 79.4 |
Performance by hardware and tool
Every hardware/tool/quantization combination Flint_t1p95k20m0 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 |
|---|---|---|---|---|---|---|---|---|
| Tesla V100-PCIE-32GB | llama.cpp | — | 90.7 tok/s | 88.9 tok/s | 58.0 GB | 32,768 tokens | 75.5 | 2 |
Benchmark runs
All 2 Flint_t1p95k20m0 runs submitted so far — expand one for its hardware, quality breakdown and per-scenario detail.
Frequently asked questions
- Is Flint_t1p95k20m0 good for coding?
- In our benchmarks, Flint_t1p95k20m0 scores 70.0/100 for coding. It runs at about 88.9 tok/s, so if you want more speed, unsloth/Qwen3.8-27B-GGUF:IQ3_S is faster (~104.3 tok/s) and still scores well for coding (83.0/100).
- Is Flint_t1p95k20m0 good for agentic (tool-using) tasks?
- In our benchmarks, Flint_t1p95k20m0 scores 78.6/100 for agentic workflows. It runs at about 88.9 tok/s, so if you want more speed, Ornith-1.5-35B-A3B-BigBang-oQ8e-mtp is faster (~102.7 tok/s) and still scores well for agentic workflows (91.1/100).
- How fast is Flint_t1p95k20m0 for local inference?
- Across 2 community benchmark runs, Flint_t1p95k20m0 reaches up to 90.7 tok/s and averages 88.9 tok/s, with the fastest results on Tesla V100-PCIE-32GB.
- How much memory does Flint_t1p95k20m0 need?
- The leanest observed configuration used about 58.0 GB of memory.
- Which tools have been used to run Flint_t1p95k20m0?
- Benchmarks were submitted using llama.cpp. Results are community-contributed and updated as new runs arrive.