Model benchmarks
LFM2.5-8B-A1B-MLX-8bit local LLM performance
As of August 2026, LFM2.5-8B-A1B-MLX-8bit runs at up to 197.3 tok/s for local inference (best of 2 community benchmark runs across 1 GPU).
Model size
8B
Peak speed
197.3 tok/s
Average speed
193.4 tok/s
Min memory
10.2 GB
Max context
65,536 tokens
Avg runtime / run
1m 30s
Avg quality
48.0
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 | 46.0 | 48.0 | 50.0 |
| Agent Workflow | 64.7 | 65.3 | 65.9 |
| Code Generation | 7.7 | 9.9 | 12.0 |
| Role Play & Narrative | 67.9 | 68.0 | 68.2 |
| Research & Analysis | 43.8 | 48.9 | 54.0 |
Performance by hardware and tool
Every hardware/tool/quantization combination LFM2.5-8B-A1B-MLX-8bit has been benchmarked on, ranked by peak token generation speed. Last updated August 2026.
| Hardware | Tool | Quant | Best tok/s | Avg tok/s | Memory | Context | Quality | Runs |
|---|---|---|---|---|---|---|---|---|
| Apple M5 Max | oMLX | — | 197.3 tok/s | 193.4 tok/s | 10.2 GB | 65,536 tokens | 48.0 | 2 |
Benchmark runs
All 2 LFM2.5-8B-A1B-MLX-8bit runs submitted so far — expand one for its hardware, quality breakdown and per-scenario detail.
Frequently asked questions
- How fast is LFM2.5-8B-A1B-MLX-8bit for local inference?
- Across 2 community benchmark runs, LFM2.5-8B-A1B-MLX-8bit reaches up to 197.3 tok/s and averages 193.4 tok/s, with the fastest results on Apple M5 Max.
- How much memory does LFM2.5-8B-A1B-MLX-8bit need?
- The leanest observed configuration used about 10.2 GB of memory.
- Which tools have been used to run LFM2.5-8B-A1B-MLX-8bit?
- Benchmarks were submitted using oMLX. Results are community-contributed and updated as new runs arrive.