BENCHMARK — TOR BRIDGE TRANSPORTS 2026
BENCHMARK · CENSORSHIP · DATA
Which Bridge Should You Use in 2026?
We benchmarked Tor's two mainstream pluggable transports — obfs4 and Snowflake — across repeated connection tests during August 2026: time-to-bootstrap, page-load latency on onion services, and resistance to network-level blocking.
TABLE OF CONTENTS
1. Methodology
- Tor Browser 13.5, default settings + bridges, fresh identity per run.
- n=60 bootstrap runs per transport over 14 days, mixed home/ISP networks (US/EU).
- Metric set: time to full bootstrap, first .onion page load, sustained throughput (10 MB fetch), bridge failure rate.
2. Results: Bootstrap & Latency
| METRIC | obfs4 | Snowflake |
|---|---|---|
| Median bootstrap | 11.8 s | 24.5 s |
| P95 bootstrap | 19.2 s | 58.7 s |
| First .onion load | 2.1 s | 4.9 s |
| Throughput (10 MB) | ~2.4 MB/s | ~0.7 MB/s |
| Run failures | 3 / 60 | 8 / 60 |
obfs4 wins on raw performance in every measured category — expected, since it proxies directly to a relay while Snowflake hops through volunteer WebRTC peers.
3. Blocking Resistance
Performance is only half the story. In networks with aggressive DPI:
- obfs4 is occasionally fingerprinted via packet-size/timing heuristics; public bridges from the built-in list degrade fastest once blocked.
- Snowflake survives DPI better because traffic masquerades as WebRTC to high-entropy volunteer endpoints; it kept working where obfs4 public bridges were throttled.
- Best practice: request private obfs4 bridges by email (
bridges@torproject.org) — private bridges avoid the public-list blocking problem entirely.
4. Verdict by Scenario
| SCENARIO | RECOMMENDED | WHY |
|---|---|---|
| Normal network, privacy-focused | No bridges needed | Bridges add latency without benefit |
| Casual filtering / blocked Tor | obfs4 (private) | Fastest stable tunnel |
| Heavy DPI / national firewall | Snowflake fallback | Harder to block wholesale |
| High-risk sessions | Private obfs4 + Tails | See Tails section |
⚠ Disclaimer: Educational and research purposes only. Benchmarks reflect our test networks; your results will vary by region and ISP.
Related reading: Tor OpSec Hardening Guide · Nexus Official Onion Links · Darknet Safety & Privacy