Caveat: tool-calling emits the correct tool and arguments, but the closing <|tool_call_end|> tag may be missing — not recommended for tool-calling runtimes. Vision/text generation is validated.

	Size	vs original
Original BF16	6.25 GB	—
This checkpoint (BF16)	4.57 GB	−27%

Quality vs base (quick suite): text/mm PPL ratio 0.89/0.82 (PASS), kld 0.33/0.37, generation chrF 0.52/0.46, integrity index 41.1 (record-only).

Load

python

from transformers import AutoModelForImageTextToText
model = AutoModelForImageTextToText.from_pretrained(
    "konic-labs/LFM2.5-VL-3B-ffn6144", dtype="bfloat16")

Derivative of LiquidAI/LFM2.5-VL-3B (LFM Open License v1.0). Pipeline: https://github.com/konic-labs/lfm25-vl-3b-lab · Collection: LFM2.5 3B Vision - Optimized.