Unified MemoryBenchmarks / Which Small AI Model Should You Run on a Mac? We Ran 312 Tests / Models
Spark-X2.5 4B
XHToken/Spark-X2.5-4B-GGUF:Q4_K_M · llama.cpp · llama-server version: 0.7.0 (build 11669, commit 620d2225c)
#4 of 8
overall, text jobs
83.1
quality score
65.6
median tokens/s
3.79 GB
RAM with 8K context filled
2.42 GB
download size
Score per job
| Job | Accuracy | Pass | Easy | Medium | Hard | Median latency | Tok/s | Job time | Errors |
|---|---|---|---|---|---|---|---|---|---|
| Assistant actions | 79.3 | 72.0% | 86.7% | 70.6% | 61.1% | 0.52 s | 67.4 | 34.8 s | 0 |
| Ask my documents | 70.9 | 62.5% | 66.7% | 76.9% | 46.7% | 2.19 s | 68.3 | 1m 36s | 0 |
| Rewrite and proofread | 88.0 | 45.2% | 64.3% | 57.1% | 14.3% | 1.76 s | 63.0 | 1m 18s | 0 |
| Summarize + extract | 88.0 | 54.5% | 66.7% | 53.8% | 43.8% | 2.35 s | 65.1 | 1m 51s | 0 |
| Translate | 84.7 | 70.0% | 92.3% | 76.9% | 42.9% | 1.29 s | 65.6 | 51.6 s | 0 |
| Sort and triage | 87.5 | 83.9% | 82.4% | 94.7% | 75.0% | 0.45 s | 72.7 | 24.5 s | 0 |
Easy / medium / hard columns are pass rates within each difficulty tier.
Memory and loading
| Download size | 2.42 GB |
| Loaded, idle | 3.42 GB |
| In use (8K context) | 3.56 GB |
| 8K context filled | 3.79 GB |
| Server peak during the jobs (incl. reserved context) | 14.98 GB |
| Load time | 54.7 s |
Measured on Apple M4 Max, 36 GB, macOS 27.0.1. How memory was measured.