Model benchmarks
LFM2.5-24B-A2B-MLX-8bit local LLM performance
As of October 2026, LFM2.5-24B-A2B-MLX-8bit runs at up to 127.1 tok/s for local inference (best of 3 community benchmark runs across 1 GPU).
Model size
24B
Peak speed
127.1 tok/s
Average speed
120.1 tok/s
Avg PP
3736.8 tok/s
Min memory
24.2 GB
Max context
262,144 tokens
Avg output / run
35,472 tokens
Avg runtime / run
5m 5s
Avg quality
63.4
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 | 59.5 | 63.4 | 66.9 |
| Agent Workflow | 15.4 | 38.8 | 53.9 |
| Code Generation | 55.1 | 62.4 | 69.3 |
| Role Play & Narrative | 72.7 | 74.9 | 78.3 |
| Research & Analysis | 75.8 | 77.6 | 78.9 |
Performance by hardware and tool
Every hardware/tool/quantization combination LFM2.5-24B-A2B-MLX-8bit 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 |
|---|---|---|---|---|---|---|---|---|
| Apple M5 Max | oMLX | 8bit | 127.1 tok/s | 120.1 tok/s | 24.2 GB | 262,144 tokens | 63.4 | 3 |
Benchmark runs
All 3 LFM2.5-24B-A2B-MLX-8bit runs submitted so far — expand one for its hardware, quality breakdown and per-scenario detail.
Frequently asked questions
- Is LFM2.5-24B-A2B-MLX-8bit good for coding?
- In our benchmarks, LFM2.5-24B-A2B-MLX-8bit scores 62.4/100 for coding. It runs at about 120.1 tok/s, so if you want more speed, Qwen3.6-35B-A3B-Q4_K_XL is faster (~156.9 tok/s) and still scores well for coding (82.2/100).
- Is LFM2.5-24B-A2B-MLX-8bit good for agentic (tool-using) tasks?
- In our benchmarks, LFM2.5-24B-A2B-MLX-8bit scores 38.8/100 for agentic workflows. It runs at about 120.1 tok/s, so if you want more speed, Tiel-Coder-35B-A3B-Q4_K_S is faster (~162.4 tok/s) and still scores well for agentic workflows (87.8/100).
- How fast is LFM2.5-24B-A2B-MLX-8bit for local inference?
- Across 3 community benchmark runs, LFM2.5-24B-A2B-MLX-8bit reaches up to 127.1 tok/s and averages 120.1 tok/s, with the fastest results on Apple M5 Max.
- How much memory does LFM2.5-24B-A2B-MLX-8bit need?
- The leanest observed configuration used about 24.2 GB of memory (quantizations tested: 8bit).
- Which tools have been used to run LFM2.5-24B-A2B-MLX-8bit?
- Benchmarks were submitted using oMLX. Results are community-contributed and updated as new runs arrive.