Unified MemoryBenchmarks / Which Small AI Model Should You Run on a Mac? We Ran 312 Tests / Models
LFM2.5-VL 3B
LiquidAI/LFM2.5-VL-3B-GGUF:Q4_K_M · llama.cpp · llama-server version: 0.7.0 (build 11669, commit 620d2225c)
#1 of 3
reading images
85.0
quality score
114.3
median tokens/s
2.81 GB
RAM with 8K context filled
2.10 GB
download size
Score per job
| Job | Accuracy | Pass | Easy | Medium | Hard | Median latency | Tok/s | Job time | Errors |
|---|---|---|---|---|---|---|---|---|---|
| Read images | 85.0 | 85.0% | 100.0% | 80.0% | 76.9% | 2.45 s | 114.3 | 1m 22s | 0 |
Easy / medium / hard columns are pass rates within each difficulty tier.
Memory and loading
| Download size | 2.10 GB |
| Loaded, idle | 2.74 GB |
| In use (8K context) | 2.77 GB |
| 8K context filled | 2.81 GB |
| Server peak during the jobs (incl. reserved context) | 4.17 GB |
| Load time | 48.5 s |
Measured on Apple M4 Max, 36 GB, macOS 27.0.1. How memory was measured.