Build snapshot
Saved Jun 8, 2026 · Rules v0.1.0 · Quant Q4_K_M
Immutable snapshot — prices and compatibility frozen at save time
Profile
Local Dev
Target model
Llama 3.3 70B
Snapshot total
$1,828.93
Llama 3.3 70B (Q4_K_M) requires ~44.5 GB VRAM but this build has 24 GB.
Compatibility checks are advisory. Verify fit, power, and cooling before purchase.
Snapshot prices frozen at save timeSAMPLE
Unable to estimate rating.
For lighter local inference, MacBook Pro 14" M4 Pro (48GB) ($2,499) runs ~42 tok/s on Llama 8B with MLX — no GPU drivers required.
Break-even at ~26 months ($71.20/mo)
Cloud rental likely cheaper
Cloud pricing is indicative. See affiliate disclosure.
| Part | Qty | Price | |------|-----|-------| | NVIDIA GeForce RTX 3090 (Used) | 1 | $546 | | AMD Ryzen 7 7800X3D | 1 | $346 | | MSI MAG B650 TOMAHAWK WIFI | 1 | $220 | | G.Skill Trident Z5 64GB DDR5-6000 | 1 | $170 | | WD Black SN850X 1TB NVMe | 1 | $105 | | Corsair RM1000x 1000W | 1 | $165 | | Lian Li Lancool 216 | 1 | $93 | | NZXT Kraken X73 360mm AIO | 1 | $184 | **Total:** $1,829 **Profile:** LOCAL_DEV **Model:** llama-3.3-70b **Compatibility:** red **Rating:** 15.5 tok/s decode **Range:**…