Downloads · 30 days
11
31% of all-time downloads
LFM2.5-VL-3B-ffn7168-int4
konic-labs/LFM2.5-VL-3B-ffn7168-int4
LFM2.5-VL-3B-ffn7168-int4 is a machine learning model from konic-labs. Use it for the machine learning task on the model card, and read the license before you ship it in a product. The card lists the license as other.
Bonus tier of the Konic surgical-compression series: FFN pruned 10752 → 7168 + distillation recovery, then native INT4/INT8 (compressed-tensors).
All-time downloads
36
Public
Parameters
3B
2.5 GB on disk
Likes
0
Public
Repo makeup
Click a slice to open those files.
.safetensors2.5 GB · 99%
Parameter types
How the weights are stored.
I322.3B · 78%
Model card
From the Hugging Face model README
LFM2.5-VL-3B-ffn7168-int4
Bonus tier of the Konic surgical-compression series: FFN pruned 10752 → 7168
- distillation recovery, then native INT4/INT8 (compressed-tensors).
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 (INT4/INT8) | 2.50 GB | −60% |
Quality vs base (quick suite): text/mm PPL ratio 0.67/0.80 (PASS), kld 0.30/0.31, generation chrF 0.46/0.49, integrity index 41.2 (record-only).
Serve with vLLM
from vllm import LLM
llm = LLM(model="konic-labs/LFM2.5-VL-3B-ffn7168-int4",
quantization="compressed-tensors")
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.
- Type
- lfm2_vl
- License
- other
- Created
- Aug 13, 2026
- Updated
- Aug 13, 2026