Model benchmarks

LFM3-Omni-2B-Q4_K_M local LLM performance

As of November 2026, LFM3-Omni-2B-Q4_K_M runs at up to 150.8 tok/s for local inference (best of 2 community benchmark runs across 1 GPU).

llama.cppQ4_K_M
ShareRedditX

Model size

2B

Peak speed

150.8 tok/s

Average speed

150.5 tok/s

Avg PP

2256.6 tok/s

Min memory

3.5 GB

Max context

131,072 tokens

Avg output / run

6,088 tokens

Avg runtime / run

1m 1s

Avg quality

36.6

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.

TaskP5 (low)AvgP95 (high)
Overall34.236.639.0
Agent Workflow66.368.069.7
Code Generation0.91.72.4
Role Play & Narrative30.638.346.0
Research & Analysis35.838.541.1

Performance by hardware and tool

Every hardware/tool/quantization combination LFM3-Omni-2B-Q4_K_M has been benchmarked on, ranked by peak token generation speed. Last updated November 2026.

HardwareToolQuantBest tok/sAvg tok/sMemoryContextQualityRuns
Advanced Micro Devices, Inc. [AMD/ATI] HawkPoint1llama.cppQ4_K_M150.8 tok/s150.5 tok/s3.5 GB131,072 tokens36.62

Benchmark runs

All 2 LFM3-Omni-2B-Q4_K_M runs submitted so far — expand one for its hardware, quality breakdown and per-scenario detail.

Frequently asked questions

Is LFM3-Omni-2B-Q4_K_M good for agentic (tool-using) tasks?
In our benchmarks, LFM3-Omni-2B-Q4_K_M scores 68.0/100 for agentic workflows. It runs at about 150.5 tok/s, so if you want more speed, unsloth/Qwen3.6-35B-A3B-GGUF:UD-Q4_K_M is faster (~246.8 tok/s) and still scores well for agentic workflows (80.6/100).
How fast is LFM3-Omni-2B-Q4_K_M for local inference?
Across 2 community benchmark runs, LFM3-Omni-2B-Q4_K_M reaches up to 150.8 tok/s and averages 150.5 tok/s, with the fastest results on Advanced Micro Devices, Inc. [AMD/ATI] HawkPoint1.
How much memory does LFM3-Omni-2B-Q4_K_M need?
The leanest observed configuration used about 3.5 GB of memory (quantizations tested: Q4_K_M).
Which tools have been used to run LFM3-Omni-2B-Q4_K_M?
Benchmarks were submitted using llama.cpp. Results are community-contributed and updated as new runs arrive.