Model benchmarks
GRM-3.2-Sky-ONYX-quality local LLM performance
As of October 2026, GRM-3.2-Sky-ONYX-quality runs at up to 12.3 tok/s for local inference (best of 15 community benchmark runs across 1 GPU).
Model size
Unknown
Peak speed
12.3 tok/s
Average speed
11.3 tok/s
Avg PP
53.2 tok/s
Min memory
n/a
Max context
262,144 tokens
Avg output / run
15,665 tokens
Avg runtime / run
26m 46s
Avg quality
78.0
Benchmark runs
15
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 15 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 | 68.9 | 78.0 | 82.6 |
| Agent Workflow | 65.9 | 76.6 | 87.7 |
| Code Generation | 64.7 | 73.3 | 78.9 |
| Role Play & Narrative | 74.9 | 85.8 | 92.5 |
| Research & Analysis | 52.3 | 76.3 | 84.8 |
Performance by hardware and tool
Every hardware/tool/quantization combination GRM-3.2-Sky-ONYX-quality 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 |
|---|---|---|---|---|---|---|---|---|
| Intel Graphics | ik_llama.cpp:cpu | — | 12.3 tok/s | 11.2 tok/s | n/a | 262,144 tokens | 79.5 | 14 |
| Intel Graphics | llama.cpp | — | 12.1 tok/s | 12.1 tok/s | n/a | 262,144 tokens | 56.4 | 1 |
Benchmark runs
All 15 GRM-3.2-Sky-ONYX-quality runs submitted so far — expand one for its hardware, quality breakdown and per-scenario detail.
Frequently asked questions
- Is GRM-3.2-Sky-ONYX-quality good for coding?
- In our benchmarks, GRM-3.2-Sky-ONYX-quality scores 73.3/100 for coding. It runs at about 11.3 tok/s, so if you want more speed, Qwen3.8-27B-Cold-Fusion-GAIN-V1.1-NM-DAU-NEO-MAX-MTP-GGUF-Q6_K is faster (~34.2 tok/s) and still scores well for coding (85.9/100).
- Is GRM-3.2-Sky-ONYX-quality good for agentic (tool-using) tasks?
- In our benchmarks, GRM-3.2-Sky-ONYX-quality scores 76.6/100 for agentic workflows. It runs at about 11.3 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 GRM-3.2-Sky-ONYX-quality for local inference?
- Across 15 community benchmark runs, GRM-3.2-Sky-ONYX-quality reaches up to 12.3 tok/s and averages 11.3 tok/s, with the fastest results on Intel Graphics.
- Which tools have been used to run GRM-3.2-Sky-ONYX-quality?
- Benchmarks were submitted using ik_llama.cpp:cpu, llama.cpp. Results are community-contributed and updated as new runs arrive.