Model benchmarks
LFM3-Omni-2B-BF16 local LLM performance
As of November 2026, LFM3-Omni-2B-BF16 runs at up to 84.9 tok/s for local inference (best of 2 community benchmark runs across 2 GPUs).
Model size
2B
Peak speed
84.9 tok/s
Average speed
66.5 tok/s
Avg PP
2766.8 tok/s
Min memory
4.7 GB
Max context
131,072 tokens
Avg output / run
4,864 tokens
Avg runtime / run
1m 20s
Avg quality
37.8
Benchmark runs
2
GPUs tested
2
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 | 29.0 | 37.8 | 46.6 |
| Agent Workflow | 61.0 | 73.3 | 85.5 |
| Code Generation | 0.1 | 0.8 | 1.4 |
| Role Play & Narrative | 34.1 | 48.0 | 61.9 |
| Research & Analysis | 20.8 | 29.2 | 37.6 |
Performance by hardware and tool
Every hardware/tool/quantization combination LFM3-Omni-2B-BF16 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 |
|---|---|---|---|---|---|---|---|---|
| NVIDIA GeForce RTX 3060 | llama.cpp | — | 84.9 tok/s | 84.9 tok/s | 4.7 GB | 32,768 tokens | 47.6 | 1 |
| Advanced Micro Devices, Inc. [AMD/ATI] HawkPoint1 | llama.cpp | — | 48.1 tok/s | 48.1 tok/s | 5.9 GB | 131,072 tokens | 28.0 | 1 |
Benchmark runs
All 2 LFM3-Omni-2B-BF16 runs submitted so far — expand one for its hardware, quality breakdown and per-scenario detail.
Frequently asked questions
- Is LFM3-Omni-2B-BF16 good for coding?
- In our benchmarks, LFM3-Omni-2B-BF16 scores 0.8/100 for coding. It runs at about 66.5 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 LFM3-Omni-2B-BF16 good for agentic (tool-using) tasks?
- In our benchmarks, LFM3-Omni-2B-BF16 scores 73.3/100 for agentic workflows. It runs at about 66.5 tok/s, so if you want more speed, Nex-N2.5-mini-APEX-Mini is faster (~88.2 tok/s) and still scores well for agentic workflows (91.3/100).
- How fast is LFM3-Omni-2B-BF16 for local inference?
- Across 2 community benchmark runs, LFM3-Omni-2B-BF16 reaches up to 84.9 tok/s and averages 66.5 tok/s, with the fastest results on NVIDIA GeForce RTX 3060.
- How much memory does LFM3-Omni-2B-BF16 need?
- The leanest observed configuration used about 4.7 GB of memory.
- Which tools have been used to run LFM3-Omni-2B-BF16?
- Benchmarks were submitted using llama.cpp. Results are community-contributed and updated as new runs arrive.