← Akhilesh's Notepad

04 / Intel LFM

LFM2.5-2.6B on Intel

OpenVINO export, deployment, and CPU/GPU/NPU profiling on the Intel Core Ultra X7 358H machine in Boston.

Status

Working end to end. Liquid Agent currently runs LFM2.5-2.6B as symmetric INT4, group size 128, on the integrated Intel GPU.

Machine and stack

Intel machine and software configuration
LayerCurrent configuration
CPUIntel Core Ultra X7 358H
GPUIntel integrated graphics 0xb080
NPUIntel AI Boost
Memory30 GiB system RAM + 8 GiB swap
OSUbuntu 26.04.1, Linux 7.0.0-30
RuntimeOpenVINO 2026.2.1 + OpenVINO GenAI 2026.2.1.0
NPU driverIntel user-mode driver 1.35.0

Inference profile

One warm-up followed by three measured runs for each device and export. Each run used 128 input tokens and generated 32 tokens. Values are median output tokens per second.

Median output tokens per second by device and model precision
DeviceFP16INT8INT4
CPU8.5 tok/s29.3 tok/s40.5 tok/s
GPU17.3 tok/s30.2 tok/s45.5 tok/s
NPU14.3 tok/s0.12 tok/s23.4 tok/s

Deployment comparison

INT4 deployment comparison across Intel devices
ConfigurationOutput speedTTFTPeak host RSSFirst compile
GPU INT445.5 tok/s188 ms1.91 GB1.3 s
CPU INT440.5 tok/s2,789 ms3.51 GB1.4 s
NPU INT423.4 tok/s925 ms3.09 GB87.9 s

Decision

GPU INT4 is the deployment choice: fastest decode, low time to first token, and lowest observed host memory.

CPU INT4 has competitive decode speed but a much slower first token. NPU INT4 works, but reaches about half the GPU's generation speed and has expensive initial compilation. The asymmetric INT8 export is pathologically slow on the NPU and should not be deployed there.

These are short-context decode measurements. Exact 16K and 32K context runs passed earlier, but long-context prefill latency and power efficiency remain separate profiling tasks.

Installed layout

Installed runtime and model locations
ComponentLocation
Runtime helper/home/liquid/intel-lfm-openvino
Native checkpoint/home/liquid/models/LFM2.5-2.6B-native
OpenVINO exports/home/liquid/models/LFM2.5-2.6B-openvino
Liquid Agent/home/liquid/liquid-context
Model serviceliquid-lfm-openvino.service on loopback port 8082
Agent bridgeliquid-agent-bridge.service on loopback port 8087

Verified

Next

  1. Upstream runtime code, scripts, service definitions, tests, and benchmark records into liquid-context.
  2. Add a model card and checksums to the exported OpenVINO model folder.
  3. Upload the three OpenVINO variants to the LiquidAI Hugging Face repository.
  4. Profile 16K/32K prefill and power efficiency separately.