Model benchmarks
Nanbeige4.3_t1p95k0m0 local LLM performance
As of November 2026, Nanbeige4.3_t1p95k0m0 runs at up to 31.1 tok/s for local inference (best of 2 community benchmark runs across 1 GPU).
Model size
33B
Peak speed
31.1 tok/s
Average speed
30.9 tok/s
Avg PP
679.7 tok/s
Min memory
51.7 GB
Max context
32,768 tokens
Avg output / run
37,749 tokens
Avg runtime / run
20m 46s
Avg quality
72.5
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 | 71.9 | 72.5 | 73.0 |
| Agent Workflow | 40.4 | 63.9 | 87.3 |
| Code Generation | 30.1 | 52.0 | 73.9 |
| Role Play & Narrative | 89.8 | 90.7 | 91.6 |
| Research & Analysis | 82.9 | 83.3 | 83.7 |
Performance by hardware and tool
Every hardware/tool/quantization combination Nanbeige4.3_t1p95k0m0 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 | — | 31.1 tok/s | 30.9 tok/s | 51.7 GB | 32,768 tokens | 72.5 | 2 |
Benchmark runs
All 2 Nanbeige4.3_t1p95k0m0 runs submitted so far — expand one for its hardware, quality breakdown and per-scenario detail.
Frequently asked questions
- Is Nanbeige4.3_t1p95k0m0 good for coding?
- In our benchmarks, Nanbeige4.3_t1p95k0m0 scores 52.0/100 for coding. It runs at about 30.9 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 Nanbeige4.3_t1p95k0m0 good for agentic (tool-using) tasks?
- In our benchmarks, Nanbeige4.3_t1p95k0m0 scores 63.9/100 for agentic workflows. It runs at about 30.9 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 Nanbeige4.3_t1p95k0m0 for local inference?
- Across 2 community benchmark runs, Nanbeige4.3_t1p95k0m0 reaches up to 31.1 tok/s and averages 30.9 tok/s, with the fastest results on Tesla V100-PCIE-32GB.
- How much memory does Nanbeige4.3_t1p95k0m0 need?
- The leanest observed configuration used about 51.7 GB of memory.
- Which tools have been used to run Nanbeige4.3_t1p95k0m0?
- Benchmarks were submitted using llama.cpp. Results are community-contributed and updated as new runs arrive.