How fast is unsloth/LFM2.5-230M-GGUF for local inference?

Across 8 community benchmark runs, unsloth/LFM2.5-230M-GGUF reaches up to 905.2 tok/s and averages 492.4 tok/s, with the fastest results on NVIDIA GeForce RTX 4070.

How much memory does unsloth/LFM2.5-230M-GGUF need?

The leanest observed configuration used about 1.6 GB of memory (quantizations tested: Q4_K_M, UD-Q4_K_XL, UD-Q8_K_XL).

Which tools have been used to run unsloth/LFM2.5-230M-GGUF?

Benchmarks were submitted using Unsloth Studio, llama.cpp. Results are community-contributed and updated as new runs arrive.

Model benchmarks

unsloth/LFM2.5-230M-GGUF local LLM performance

As of October 2026, unsloth/LFM2.5-230M-GGUF runs at up to 905.2 tok/s for local inference (best of 8 community benchmark runs across 2 GPUs).

Unsloth Studiollama.cppQ4_K_MUD-Q4_K_XLUD-Q8_K_XL

ShareReddit X

Model size

230M

Peak speed

905.2 tok/s

Average speed

492.4 tok/s

Avg PP

3143.8 tok/s

Min memory

1.6 GB

Max context

128,000 tokens

Avg output / run

4,151 tokens

Avg runtime / run

14s

Avg quality

13.7

Benchmark runs

GPUs tested

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 8 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	11.5	13.7	16.6
Agent Workflow	9.9	20.7	29.2
Code Generation	0.0	0.3	1.6
Role Play & Narrative	13.0	18.7	23.8
Research & Analysis	6.7	15.0	21.2

Performance by hardware and tool

Every hardware/tool/quantization combination unsloth/LFM2.5-230M-GGUF 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
NVIDIA GeForce RTX 4070	Unsloth Studio	Q4_K_M	905.2 tok/s	905.2 tok/s	1.8 GB	65,536 tokens	16.4	1
NVIDIA GeForce RTX 4070	Unsloth Studio	UD-Q4_K_XL	900.8 tok/s	900.8 tok/s	1.6 GB	65,536 tokens	12.2	1
NVIDIA GeForce RTX 4070	Unsloth Studio	UD-Q8_K_XL	688.8 tok/s	688.8 tok/s	2.1 GB	65,536 tokens	13.5	1
Intel(R) Arc(TM) 140V GPU (16GB)	llama.cpp	—	359.5 tok/s	303.5 tok/s	n/a	128,000 tokens	13.9	3
Intel(R) Arc(TM) 140V GPU (16GB)	Unsloth Studio	—	272.5 tok/s	266.8 tok/s	n/a	128,000 tokens	12.9	2

Benchmark runs

All 8 unsloth/LFM2.5-230M-GGUF runs submitted so far — expand one for its hardware, quality breakdown and per-scenario detail.

Open run page →

Frequently asked questions

How fast is unsloth/LFM2.5-230M-GGUF for local inference?: Across 8 community benchmark runs, unsloth/LFM2.5-230M-GGUF reaches up to 905.2 tok/s and averages 492.4 tok/s, with the fastest results on NVIDIA GeForce RTX 4070.
How much memory does unsloth/LFM2.5-230M-GGUF need?: The leanest observed configuration used about 1.6 GB of memory (quantizations tested: Q4_K_M, UD-Q4_K_XL, UD-Q8_K_XL).
Which tools have been used to run unsloth/LFM2.5-230M-GGUF?: Benchmarks were submitted using Unsloth Studio, llama.cpp. Results are community-contributed and updated as new runs arrive.