This page mixes current runtime metrics and deeper R&D measurements. Start with the reproduction commands and table notes before comparing figures directly.
| Role | Provider | CPU | Memory |
|---|---|---|---|
| Benchmark server | Hetzner AX102 | AMD Ryzen 9 7950X3D | 128 GB DDR5 |
| Demo gateway | Helsinki | AMD EPYC 9454 | 128 GB DDR5 |
| Training | Vast.ai | NVIDIA A100 80 GB | -- |
| Operation | Requests | Errors | p50 | p99 |
|---|---|---|---|---|
| Cached execution (.syn) | 1,000,000 | 0 | 0.004ms | 0.013ms |
| Cold start (.syn) | -- | -- | 0.063ms | 0.539ms |
| Python transpile (cold) | -- | -- | 0.678ms | -- |
| Python transpile (cached) | 1,000 | 0 | 0.005ms | 0.015ms |
Cached means the compiled Wasm module is reused from the moka cache. Cold starts compile from source. Real-world AI agent traffic is a mix of both.
| Metric | Synapse (measured) | AWS Lambda | E2B | Cloudflare Workers |
|---|---|---|---|---|
| Cold Start | <2ms | 200-500ms * | ~200ms * | <5ms * |
| Memory / Sandbox | 64KB | 128MB min * | ~150MB * | 128MB * |
| Sandboxes / $60 Server | ~130,000 (theoretical) | N/A | N/A | N/A |
| Isolation | Wasm Sandbox | Firecracker | Firecracker | V8 Isolate |
* Competitor figures sourced from their published documentation as of March 2026. Sandbox density is a theoretical calculation based on 64KB arena size — not a tested concurrent load. Cost comparison removed pending independent verification.
| Workload | Docker | Synapse | Speed | Energy |
|---|---|---|---|---|
| noop | 233ms | 0.015ms | 15,533x | 7.4x |
| arithmetic | 246ms | 0.025ms | 9,840x | 7.9x |
| fibonacci | 321ms | 2.261ms | 142x | 9.1x |
| prime_sieve | 251ms | 0.721ms | 348x | 7.6x |
| memory_stress | 248ms | 0.033ms | 7,515x | 8.0x |
All measurements from real runs on Hetzner AX102 (AMD Ryzen 9 7950X3D, 128 GB DDR5).
Measured via RAPL (Running Average Power Limit) instrumentation on AMD EPYC.
SYNAPSE_START_GATEWAY=1 python3 tests/release_test.py SYNAPSE_START_GATEWAY=1 python3 tests/test_gateway_verification_http.py SYNAPSE_START_GATEWAY=1 python3 tests/test_self_hosted_examples.py python3 benchmarks/eval_only_benchmark.py # Extended energy study on benchmark hardware bash benchmarks/energy_benchmark.sh --iterations 1000
The first four commands are the current product proof set. The energy benchmark is a deeper methodology pass for the benchmark server, not a prerequisite for evaluating the runtime locally.