title string | subtitle string | date timestamp[s] | hardware dict | device_specs list | live_benchmarks list | known_benchmarks list | total_devices_connected int64 | cost string |
|---|---|---|---|---|---|---|---|---|
dispatchAI Real-Device Silicon Profiling Report | Snapdragon 865 (Kryo 585) — Samsung S20 FE 5G | 2026-06-26T00:00:00 | {
"chipset": "Qualcomm Snapdragon 865 (SM8250)",
"codename": "kona",
"process": "7nm TSMC",
"cpu": "Kryo 585: 1x2.84GHz + 3x2.42GHz + 4x1.80GHz",
"gpu": "Adreno 650",
"npu": "Hexagon Tensor Accelerator",
"memory": "8GB LPDDR5 (7.47GB total, 3-3.7GB available)",
"device": "Samsung Galaxy S20 FE 5G (SM-G9... | [
{
"serial": "R3CN509PLHA",
"model": "SM-G981V",
"soc": "SM8250",
"cores": 9,
"ram_gb": 7.47,
"ram_free_gb": 3.7
},
{
"serial": "R3CN50A0HSV",
"model": "SM-G981V",
"soc": "SM8250",
"cores": 9,
"ram_gb": 7.47,
"ram_free_gb": 3.37
},
{
"serial": "R3CN50AXHWE"... | [
{
"serial": "R3CN509PLHA",
"model": "model.gguf (on-device)",
"generation_tps": 10.9,
"success": true,
"notes": "Live benchmark, Jun 26 2026"
}
] | [
{
"model": "SmolLM2-135M fp16",
"gen_tps": 18.3,
"prompt_tps": 85
},
{
"model": "Qwen2.5-0.5B Q4",
"gen_tps": 9.8,
"prompt_tps": 47.5
},
{
"model": "Llama-3.2-1B fp16",
"gen_tps": 6.5,
"prompt_tps": 28
},
{
"model": "TinyLlama-1.1B Q5",
"gen_tps": 6.2,
"pr... | 39 | $0 |
Real-Device Silicon Profiling: Snapdragon 865
Per-device inference benchmarks on real Samsung S20 FE 5G phones (Snapdragon 865). No simulation. Real ARM CPU inference.
Hardware
| Property | Value |
|---|---|
| Chipset | Qualcomm Snapdragon 865 (SM8250) |
| CPU | Kryo 585: 1x2.84GHz + 3x2.42GHz + 4x1.80GHz |
| GPU | Adreno 650 |
| NPU | Hexagon Tensor Accelerator |
| RAM | 8GB LPDDR5 (7.47GB total, 3-3.7GB free) |
| Device | Samsung Galaxy S20 FE 5G (SM-G981V) |
| Devices connected | 39 |
Live Benchmark (Jun 26, 2026)
| Model | Device | Gen t/s |
|---|---|---|
| model.gguf (on-device) | R3CN509PLHA | 10.9 |
Known Benchmarks (from phone farm)
| Model | Gen t/s | Prompt t/s |
|---|---|---|
| SmolLM2-135M fp16 | 18.3 | 85.0 |
| Qwen2.5-0.5B Q4 | 9.8 | 47.5 |
| Llama-3.2-1B fp16 | 6.5 | 28.0 |
| TinyLlama-1.1B Q5 | 6.2 | 26.5 |
| Qwen2.5-1.5B Q5 | 4.2 | 18.0 |
| Gemma-2-2B Q5 | 3.0 | 12.5 |
Dispatch AI (FZE), Sharjah SRTI Free Zone, License No. 10818.
- Downloads last month
- 13