Model benchmarks
granite-3.3-8b-instruct local LLM performance
As of October 2026, granite-3.3-8b-instruct runs at up to 14.1 tok/s for local inference (best of 3 community benchmark runs across 1 GPU).
Model size
8B
Peak speed
14.1 tok/s
Average speed
13.6 tok/s
Avg PP
1850.0 tok/s
Min memory
7.2 GB
Max context
32,768 tokens
Avg output / run
4,402 tokens
Avg runtime / run
5m 35s
Avg quality
49.9
Benchmark runs
3
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 3 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 | 47.8 | 49.9 | 52.6 |
| Agent Workflow | 66.9 | 72.0 | 78.1 |
| Code Generation | 11.2 | 15.8 | 18.9 |
| Role Play & Narrative | 63.0 | 64.6 | 66.5 |
| Research & Analysis | 42.6 | 47.1 | 51.6 |
Performance by hardware and tool
Every hardware/tool/quantization combination granite-3.3-8b-instruct 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 |
|---|---|---|---|---|---|---|---|---|
| NVIDIA GeForce RTX 3060 Laptop GPU | Bionic | — | 14.1 tok/s | 13.6 tok/s | 7.2 GB | 32,768 tokens | 49.9 | 3 |
Benchmark runs
All 3 granite-3.3-8b-instruct runs submitted so far — expand one for its hardware, quality breakdown and per-scenario detail.
Frequently asked questions
- Is granite-3.3-8b-instruct good for coding?
- In our benchmarks, granite-3.3-8b-instruct scores 15.8/100 for coding. It runs at about 13.6 tok/s, so if you want more speed, tensorfold-qwen38-27b is faster (~25.0 tok/s) and still scores well for coding (84.3/100).
- Is granite-3.3-8b-instruct good for agentic (tool-using) tasks?
- In our benchmarks, granite-3.3-8b-instruct scores 72.0/100 for agentic workflows. It runs at about 13.6 tok/s, so if you want more speed, muse-glimmer:latest is faster (~33.3 tok/s) and still scores well for agentic workflows (91.6/100).
- How fast is granite-3.3-8b-instruct for local inference?
- Across 3 community benchmark runs, granite-3.3-8b-instruct reaches up to 14.1 tok/s and averages 13.6 tok/s, with the fastest results on NVIDIA GeForce RTX 3060 Laptop GPU.
- How much memory does granite-3.3-8b-instruct need?
- The leanest observed configuration used about 7.2 GB of memory.
- Which tools have been used to run granite-3.3-8b-instruct?
- Benchmarks were submitted using Bionic. Results are community-contributed and updated as new runs arrive.