Model benchmarks
LFM2.5-8B-A1B-Q8_0 local LLM performance
As of November 2026, LFM2.5-8B-A1B-Q8_0 runs at up to 90.2 tok/s for local inference (best of 3 community benchmark runs across 1 GPU).
Model size
8B
Peak speed
90.2 tok/s
Average speed
89.5 tok/s
Avg PP
1179.3 tok/s
Min memory
10.8 GB
Max context
131,072 tokens
Avg output / run
17,131 tokens
Avg runtime / run
3m 22s
Avg quality
52.0
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 | 46.4 | 52.0 | 56.7 |
| Agent Workflow | 69.1 | 79.8 | 86.3 |
| Code Generation | 10.1 | 13.9 | 16.5 |
| Role Play & Narrative | 63.9 | 71.5 | 77.1 |
| Research & Analysis | 36.6 | 42.9 | 53.1 |
Performance by hardware and tool
Every hardware/tool/quantization combination LFM2.5-8B-A1B-Q8_0 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 |
|---|---|---|---|---|---|---|---|---|
| Advanced Micro Devices, Inc. [AMD/ATI] HawkPoint1 | llama.cpp | Q8_0 | 90.2 tok/s | 89.5 tok/s | 10.8 GB | 131,072 tokens | 52.0 | 3 |
Benchmark runs
All 3 LFM2.5-8B-A1B-Q8_0 runs submitted so far — expand one for its hardware, quality breakdown and per-scenario detail.
Frequently asked questions
- Is LFM2.5-8B-A1B-Q8_0 good for coding?
- In our benchmarks, LFM2.5-8B-A1B-Q8_0 scores 13.9/100 for coding. It runs at about 89.5 tok/s, so if you want more speed, ornith-1.5-35b-a3b-apex-mtp-i-quality is faster (~105.2 tok/s) and still scores well for coding (83.6/100).
- Is LFM2.5-8B-A1B-Q8_0 good for agentic (tool-using) tasks?
- In our benchmarks, LFM2.5-8B-A1B-Q8_0 scores 79.8/100 for agentic workflows. It runs at about 89.5 tok/s, so if you want more speed, ornith-1.5-35b-a3b-apex-mtp-quality is faster (~103.2 tok/s) and still scores well for agentic workflows (88.5/100).
- How fast is LFM2.5-8B-A1B-Q8_0 for local inference?
- Across 3 community benchmark runs, LFM2.5-8B-A1B-Q8_0 reaches up to 90.2 tok/s and averages 89.5 tok/s, with the fastest results on Advanced Micro Devices, Inc. [AMD/ATI] HawkPoint1.
- How much memory does LFM2.5-8B-A1B-Q8_0 need?
- The leanest observed configuration used about 10.8 GB of memory (quantizations tested: Q8_0).
- Which tools have been used to run LFM2.5-8B-A1B-Q8_0?
- Benchmarks were submitted using llama.cpp. Results are community-contributed and updated as new runs arrive.