Skip to content

How MOQOM compares

MOQOM’s own numbers are in Benchmarks, including latency by resolution, led by 4K. This page sets them beside what other platforms publish about themselves.

WebRTC platforms move big audiences to a slower path; MoQ relays forward one stream to everyone

Latency: MOQOM measured 56–57 ms against vendors’ documented figures

Headroom drawn as an Elliott wave: 2, a few dozen, 2,000 and 6,000 tested, 1 million next

  • One protocol for calls and broadcast. Built on the IETF’s Media over QUIC Transport (speaking draft-ietf-moq-transport-22 and -19): the stream that carries a two-person call also reaches an audience of thousands, without moving to a second product.
  • Conversational latency, measured end to end. 56–57 ms median glass-to-glass at 720p30 (p95 70 ms), camera to screen on two iPhone 15 Pros.*
  • Broadcast scale on modest hardware. 6,000 viewers on one 4-vCPU relay with zero loss (5.46 Gbit/s), the largest size tested.*
  • Already ahead for calls and private rooms. Conferences, private rooms and one-to-one calls hold a few dozen people at most. We tested 6,000. Glass-to-glass is 56–57 ms, under Agora’s advertised “<200 ms”, and the price is lower at every tier.*
  • The last viewer is as fast as the first, because it’s MoQ. WebRTC was built for calls, so big audiences get moved to a slower mode (Agora: 400–800 ms host to audience) or a separate product (AWS). Media over QUIC is publish and subscribe: relays forward the same media to every subscriber and chain like a CDN. Nobody gets demoted to the cheap seats.
  • Scale is built in, not bolted on. Relays shard across cores, scale out and chain to other relays. The Go control plane scales out too.
  • Encryption that costs almost nothing. Per-frame end-to-end encryption (SFrame, AES-256-GCM) at 54 µs per 1080p keyframe,* with post-quantum key agreement and signatures (ML-KEM, ML-DSA). TLS 1.3 on every connection, always.
  • Tokens bound to the device. A join token copied off one device does not work on another.
  • Your relays or ours. Self-host relays and pay only the platform fee.
  • Prepaid, 15% under Agora. 85% of Agora’s price at every tier and volume band, always at least 1¢ per 1,000 minutes under, enforced in code.
  • Open-source SDKs. Go and Rust backend SDKs and a Swift client shell are open source; the media cores ship as signed binaries.
  • Fewer platforms today. iOS and macOS are available. Web, Android, Desktop and the OBS plugin: Scoped, work ongoing.
  • No independent audit yet.
  • Not yet tested at a million viewers. So far: 6,000 on one 4-vCPU relay at 30 fps, and 2,000 watching real H.264 video at 2.78 Gbit/s, both with zero loss. A million is next. We’d rather measure it than say it.
  • The protocol is still a draft. Media over QUIC Transport is an IETF Internet-Draft, not yet an RFC. MOQOM speaks draft-22 and draft-19.
  • Post-quantum end-to-end encryption is required in every room, which is why MOQOM launches on Apple platforms only.

Competitor figures are what each vendor’s own documentation says, accessed 6 Oct 2026. Where a vendor does not document something, we say so rather than guess. No latency figure below was measured by us on the same rig as MOQOM’s, so read them as published claims, not a head-to-head test.

Documented latency Calls and broadcast
MOQOM 56–57 ms median glass-to-glass, 720p30 (measured in lab) One stream and one protocol for both
Agora “<200 ms global latency” (homepage). Docs: Interactive Live Streaming 400–800 ms host to audience; Broadcast Streaming 1,500–2,000 ms One SDK; the audience latency level affects the price
Tencent RTC (TRTC) Average end-to-end under 300 ms globally; live co-anchoring as low as 200 ms, viewing under 1 s Calls and live rooms in TRTC; CDN delivery through Cloud Streaming Services: LEB 300–1,000 ms, HTTP-FLV 2–3 s, HLS 10–30 s
AWS IVS Real-Time Under 300 ms host to viewer Up to 12 publishers per stage; 10,000 subscribers per stage by default (adjustable); publish up to 720p. Stage broadcast to an IVS channel: those viewers get low latency
AWS IVS Low-Latency Under 5 s Broadcast only
AWS Chime SDK Not documented Meetings: up to 250 participants at 720p, or 25 at 1080p; replication to 10,000 / 1,000
Google Cloud No end-to-end figure documented; Live Stream API segments are 2–20 s (default 6 s) Broadcast: Live Stream API (HLS/DASH) with Media CDN. No Google Cloud real-time video-calling SDK found; Meet Media API (Developer Preview) reads media from Google Meet calls
Cloudflare Not documented for MoQ MoQ relays (beta) on draft-14 and draft-16. Its MoQ docs list no Cloudflare client SDK for MoQ (examples use open-source moq-rs). WebRTC SDKs: RealtimeKit, Realtime SFU

Price per 1,000 HD video minutes: MOQOM $3.39, Agora and Tencent $3.99, AWS Chime $1.70, AWS IVS Real-Time about $1.20, Cloudflare free during beta

At the higher tiers MOQOM stays 15% under Agora’s and Tencent’s list prices: $7.64 Full HD, $13.59 2K and $30.59 2K+ per 1,000 minutes, against their $8.99, $15.99 and $35.99 (Tencent calls its top tier 4K). For broadcast audiences, Agora lists $1.99–$17.99 and Tencent $2.00–$18.00 per 1,000 minutes. AWS IVS Low-Latency charges per hour of output delivered: $0.036 SD, $0.072 HD and $0.144 Full HD (North America, first 10,000 hours), plus $0.20–$2.00 an hour of input. Google Cloud’s live streaming is a broadcast product and isn’t priced here.

AWS’s per-participant list prices are below MOQOM’s HD rate. Compare on your own workload, and see self-hosted relays at $0.05 per GiB.

Media encryption Protocol
MOQOM TLS 1.3 always; post-quantum SFrame E2EE (AES-256-GCM, ML-KEM, ML-DSA), required in every room IETF MoQ Transport, draft-19
Agora Built-in media encryption opt-in; E2EE in beta (web, by request) Agora SD-RTN™
Tencent RTC Built-in transport encryption by default; private payload encryption opt-in, not available with cloud recording or CDN relay UDP-based Tencent protocol; WebRTC on web
AWS IVS Real-Time Not stated on pages cited Not stated on pages cited
AWS IVS Low-Latency Not stated on pages cited RTMP, RTMPS or SRT ingest
AWS Chime SDK DTLS transport; E2EE not stated on pages cited WebRTC
Google Cloud Content encryption: AES-128, SAMPLE-AES, MPEG-CENC SRT/RTMP in, HLS/DASH out
Cloudflare Not stated on pages cited (MoQ) IETF MoQ Transport, draft-14 and draft-16; draft-18 in testing

Agora. MOQOM charges 85% of Agora’s price at every tier, after Agora’s own volume discounts, and always at least 1¢ less per 1,000 minutes; the code enforces it. Agora bills by the total pixels a user receives; MOQOM uses the same tier boundaries so the rates compare like with like.

Tencent RTC. Tencent’s video tiers and list prices match Agora’s, with the top tier named 4K. TRTC documents that, to meet data-security storage requirements, it stores account and troubleshooting data in the Chinese mainland for Tencent Cloud customers and in Singapore for Tencent Cloud International customers, and that it processes audio and video only to deliver them, without storing them. Teams with data-residency or jurisdiction requirements should check which Tencent entity and region apply to their account. MOQOM runs on Google Cloud today, and you can run relays where you choose.

Google Cloud. Google Cloud’s Live Stream API takes SRT or RTMP in and writes HLS or DASH segments of 2–20 seconds to Cloud Storage for delivery through Media CDN: a broadcast pipeline, not a real-time one. MOQOM runs on Google Cloud today.

AWS. AWS splits live video across products: IVS Real-Time (under 300 ms, up to 12 publishers, 720p publish), IVS Low-Latency (under 5 s) for large audiences, and the Chime SDK for meetings. Viewers of a stage broadcast to an IVS channel get low latency, not real-time latency. MOQOM carries the call and the broadcast on one stream and one standard.

Cloudflare. Cloudflare runs MoQ relays across its network (beta, free during the beta), so MoQ is becoming infrastructure, not one company’s protocol. Cloudflare’s MoQ documentation lists no Cloudflare client SDK for MoQ; its examples use the open-source moq-rs tools, and its client SDKs (RealtimeKit) are WebRTC. MOQOM is the SDK and platform layer. We hope to partner with Cloudflare; no partnership is signed.