Best realtime tools for web apps (search query)

We asked 10 models: best realtime chat backend for web app Each named its own top 10.

Asked 4 ways: List request · Buyer asking for advice · Search query · Worth paying for?. Share of voice pools every wording.

Agreement between models: 50%
Same model, run twice: 69%
Different picks: 21

When agreement between models is well below a model's agreement with itself, the split reflects real differences in taste rather than randomness.

The consensus

Each cell is the position that model gave the pick; blank means it left the pick off. Click a row for each model's reasoning.

#	Pick	Listed by	Claude Haiku 5.5	GPT-6 Luna	Gemini 4 Argon	Qwen 3.8 Flash	DeepSeek V4 Pro	GLM-5.3	Kimi K3	MiniMax M3	Mistral Large 4	Nemotron 3 Ultra
1	Ably	10/10	2	1	3	3	1	2	1	5	3	2
GPT-6 Luna (#1) Excellent global realtime infrastructure, reliable delivery, presence, and mature developer tooling. DeepSeek V4 Pro (#1) Purpose-built for realtime, lowest latency, rich features like presence and message history Kimi K3 (#1) Purpose-built pub/sub with global edge network, presence, history, excellent SDKs. Claude Haiku 5.5 (#2) Strong global edge network, presence, and guaranteed message ordering at scale. GLM-5.3 (#2) Enterprise-grade realtime infrastructure with excellent reliability, global edge delivery, and protocols. Nemotron 3 Ultra (#2) Highly reliable realtime platform with guaranteed delivery Gemini 4 Argon (#3) Unmatched pub/sub reliability, ordering guarantees, and global edge scalability for custom chat. Qwen 3.8 Flash (#3) Reliable pub/sub with presence and history, excellent DX. Mistral Large 4 (#3) Managed infrastructure with guaranteed ordering, presence, and 99.999% SLA. MiniMax M3 (#5) Enterprise-grade messaging with global guarantees and scaling.
2	Pusher	10/10	1	3	4	6	2	1	2	3	2	1
Claude Haiku 5.5 (#1) Hosted pub/sub with simple APIs, reliable delivery, and quick setup for web apps. GLM-5.3 (#1) Mature, developer-friendly WebSocket platform with Channels SDKs; the industry default. Nemotron 3 Ultra (#1) Easiest managed WebSocket service with global edge network DeepSeek V4 Pro (#2) Dead-simple SDKs, massive adoption, great for quick chat integration Kimi K3 (#2) Fastest to ship; dead-simple channels, presence, and client events at scale. Mistral Large 4 (#2) Simple API, excellent docs, reliable hosted WebSocket channels. GPT-6 Luna (#3) Simple, dependable hosted channels for adding live messaging to existing applications. MiniMax M3 (#3) Reliable managed WebSockets with simple pub/sub API. Gemini 4 Argon (#4) Simplest developer experience for adding real-time WebSocket events to web apps quickly. Qwen 3.8 Flash (#6) Simple pub/sub websockets for lightweight realtime features.
3	Firebase Realtime Database	10/10	6	7	7	1	3	8	4	1	4	5
Qwen 3.8 Flash (#1) Easy sync backend; fine for small chats, weak at scale. MiniMax M3 (#1) Google-backed, low-latency sync, generous free tier. DeepSeek V4 Pro (#3) Google-backed, generous free tier, seamless with other Firebase services Kimi K3 (#4) Battle-tested sync, offline support, generous free tier, minimal ops burden. Mistral Large 4 (#4) Zero backend ops, instant sync across clients, generous free tier. Nemotron 3 Ultra (#5) Zero-setup sync with offline support and auth integration Claude Haiku 5.5 (#6) Easy realtime sync for web apps, though it lacks advanced chat features. Gemini 4 Argon (#7) Tried-and-true real-time state synchronization with effortless client SDKs and Google scale.
4	Socket.IO	9/10	4		10	4	4	3	3	4	1	4
Mistral Large 4 (#1) Battle-tested Node.js library, automatic fallback, minimal setup for most apps. GLM-5.3 (#3) De-facto open-source library; full control when self-hosting Node.js WebSockets. Kimi K3 (#3) Open-source, auto-reconnect, rooms; self-host for full control and zero per-message fees. Claude Haiku 5.5 (#4) Mature, widely used Node.js library with fallbacks and a huge ecosystem. Qwen 3.8 Flash (#4) Open-source websocket layer; flexible but you host it yourself. DeepSeek V4 Pro (#4) Battle-tested, fallback transports, full control with Node.js backend MiniMax M3 (#4) Full control, flexible, but you manage scaling yourself. Nemotron 3 Ultra (#4) Most popular open-source library, self-hosted flexibility Gemini 4 Argon (#10) Industry-standard open-source library for full control over custom real-time WebSocket servers.
5	Supabase	10/10	8	6	6	5	5	4	5	2	5	3
MiniMax M3 (#2) Open-source Postgres with WebSocket subscriptions and presence. Nemotron 3 Ultra (#3) PostgreSQL-based realtime with row-level security GLM-5.3 (#4) Postgres-native realtime subscriptions; great pick for modern full-stack apps. Qwen 3.8 Flash (#5) Open Firebase alternative with postgres-backed channels. DeepSeek V4 Pro (#5) Postgres-native, open source, built-in auth and row-level security Kimi K3 (#5) Postgres-backed realtime plus auth and database in one stack; great DX. Mistral Large 4 (#5) Postgres changes as a stream, built-in auth, open source and self-hostable. GPT-6 Luna (#6) Open-source-friendly Postgres realtime stack with auth and straightforward integration. Gemini 4 Argon (#6) Great open-source Postgres backend with built-in real-time subscriptions and auth integration. Claude Haiku 5.5 (#8) Open-source Postgres-backed realtime channels, good for teams wanting self-hosting options.
6	PubNub	10/10	3	2	5	8	6	6	6	6	6	6
GPT-6 Luna (#2) Battle-tested realtime messaging with presence, access controls, and broad infrastructure reach. Claude Haiku 5.5 (#3) Global real-time data network with chat SDKs and strong uptime guarantees. Gemini 4 Argon (#5) Battle-tested global data stream network with dedicated chat SDKs and presence features. DeepSeek V4 Pro (#6) Global edge network, 99.999% SLA, robust for scale-heavy apps GLM-5.3 (#6) Massive scale pub/sub with presence and history; battle-tested at enterprise scale. Kimi K3 (#6) Enterprise-grade messaging fabric with Functions, presence, and strong compliance. MiniMax M3 (#6) Real-time network with presence, history, and low latency. Mistral Large 4 (#6) Global edge network, rich feature set, strong for high-scale messaging. Nemotron 3 Ultra (#6) Enterprise-grade messaging at massive scale
7	Stream Chat	6/10		4		2	7	7	7	8
Qwen 3.8 Flash (#2) Purpose-built chat API; fastest path to polished realtime messaging. DeepSeek V4 Pro (#7) Full chat SDK with typing indicators, threads, and rich media out of the box GLM-5.3 (#7) Polished chat-specific APIs and UI kits; fastest path to full chat features. MiniMax M3 (#8) Polished APIs, threads, reactions, and offline support.
8	Sendbird	6/10	7	5	2	7	10			7
Gemini 4 Argon (#2) Feature-complete enterprise chat platform with robust moderation, analytics, and reliability. GPT-6 Luna (#5) Comprehensive chat APIs and UI kits handle demanding messaging use cases. Claude Haiku 5.5 (#7) Purpose-built chat platform with rich features and good web SDK documentation. Qwen 3.8 Flash (#7) Enterprise-grade chat SDK with strong moderation and scale. MiniMax M3 (#7) Drop-in chat SDK with rich messaging features pre-built.
9	Stream	4/10	5	8	1				10
Gemini 4 Argon (#1) Best overall with polished UI kits, ultra-low latency, and dedicated chat infrastructure. Claude Haiku 5.5 (#5) Full chat SDKs with moderation, reactions, and polished UI kits for web. GPT-6 Luna (#8) Purpose-built chat APIs offer rich features, polished SDKs, and fast implementation. Kimi K3 (#10) Chat-specific APIs and UI kits; unread counts, threads, moderation out of box.
10	Firebase	4/10		9				5			7	8
GLM-5.3 (#5) Easiest realtime sync with Firestore listeners; ideal for quickly shipping web apps. GPT-6 Luna (#9) Fast to build with, especially for web apps already using Google's ecosystem.
11	Centrifugo	5/10	10			9	8				9	7
Nemotron 3 Ultra (#7) Language-agnostic self-hosted server, simple protocol DeepSeek V4 Pro (#8) Self-hosted, language-agnostic, excellent for owning your infrastructure Qwen 3.8 Flash (#9) Self-hostable scalable realtime server, great OSS option. Mistral Large 4 (#9) Open-source, self-hosted, supports WebSocket/SSE/HTTP with low latency. Claude Haiku 5.5 (#10) Fast open-source real-time messaging server, scalable and language-agnostic.
12	SignalR	2/10							8		8
Mistral Large 4 (#8) Native .NET support, automatic transport negotiation, solid for Microsoft stacks.
13	Amazon AppSync	4/10				10		9		10	10
GLM-5.3 (#9) Managed GraphQL subscriptions fitting AWS stacks; solid but complex setup. Qwen 3.8 Flash (#10) Managed GraphQL subscriptions for event-driven chat apps. Mistral Large 4 (#10) Serverless GraphQL subscriptions, deep AWS integration, scales automatically.
14	Convex	1/10			8
Gemini 4 Argon (#8) Modern reactive backend-as-a-service with end-to-end TypeScript and automatic real-time state sync.
15	Liveblocks	2/10		10					9
Kimi K3 (#9) Best for collaborative presence and chat UX; React hooks feel native and polished. GPT-6 Luna (#10) Developer-friendly realtime primitives and presence work well for collaborative web products.

6 lower-scoring picks are not shown; every ballot below lists them in full.

Most divisive

Picks most models listed, at very different positions

Socket.IO#1 for Mistral Large 4 #10 for Gemini 4 Argon
Stream#1 for Gemini 4 Argon #10 for Kimi K3
Sendbird#2 for Gemini 4 Argon #10 for DeepSeek V4 Pro

Hot takes

Picks that appear on only one ballot

Convex at #8, only Gemini 4 Argon
Amazon API Gateway at #9, only Claude Haiku 5.5
CometChat at #9, only Gemini 4 Argon
Amazon IVS Real-Time at #9, only MiniMax M3
Axe API at #9, only Nemotron 3 Ultra
Soketi at #10, only GLM-5.3

Every ballot

Hover a pick to find it on the other ballots. Marked picks appear on no other ballot.

Claude Haiku 5.5

Pusher
Ably
PubNub
Socket.IO
Stream
Firebase Realtime Database
Sendbird
Supabase
Amazon API Gateway
Centrifugo

67% with consensus · 92% self-consistent

GPT-6 Luna

Ably
PubNub
Pusher
Stream Chat
Sendbird
Supabase
Firebase Realtime Database
Stream
Firebase
Liveblocks

71% with consensus · 55% self-consistent

Gemini 4 Argon

Stream
Sendbird
Ably
Pusher
PubNub
Supabase
Firebase Realtime Database
Convex
CometChat
Socket.IO

41% with consensus · 92% self-consistent

Qwen 3.8 Flash

Firebase Realtime Database
Stream Chat
Ably
Socket.IO
Supabase
Pusher
Sendbird
PubNub
Centrifugo
Amazon AppSync

57% with consensus · 44% self-consistent

DeepSeek V4 Pro

Ably
Pusher
Firebase Realtime Database
Socket.IO
Supabase
PubNub
Stream Chat
Centrifugo
PartyKit
Sendbird

96% with consensus · 90% self-consistent

GLM-5.3

Pusher
Ably
Socket.IO
Supabase
Firebase
PubNub
Stream Chat
Firebase Realtime Database
Amazon AppSync
Soketi

69% with consensus · 61% self-consistent

Kimi K3

Ably
Pusher
Socket.IO
Firebase Realtime Database
Supabase
PubNub
Stream Chat
SignalR
Liveblocks
Stream

92% with consensus · 63% self-consistent

MiniMax M3

Firebase Realtime Database
Supabase
Pusher
Socket.IO
Ably
PubNub
Sendbird
Stream Chat
Amazon IVS Real-Time
Amazon AppSync

61% with consensus · 74% self-consistent

Mistral Large 4

Socket.IO
Pusher
Ably
Firebase Realtime Database
Supabase
PubNub
Firebase
SignalR
Centrifugo
Amazon AppSync

67% with consensus · 68% self-consistent

Nemotron 3 Ultra

Pusher
Ably
Supabase
Socket.IO
Firebase Realtime Database
PubNub
Centrifugo
Firebase
Axe API
PartyKit

71% with consensus · 54% self-consistent

Who agrees with whom

Overlap between two models' lists, weighted toward the top · 100 means identical · Color runs from 33 (lightest) to 92 (darkest)

	Claude Haiku 5.5	GPT-6 Luna	Gemini 4 Argon	Qwen 3.8 Flash	DeepSeek V4 Pro	GLM-5.3	Kimi K3	MiniMax M3	Mistral Large 4	Nemotron 3 Ultra
Claude Haiku 5.5		60	47	42	65	76	64	46	57	79
GPT-6 Luna	60		48	43	67	52	68	40	42	49
Gemini 4 Argon	47	48		33	38	36	38	34	38	38
Qwen 3.8 Flash	42	43	33		57	44	51	73	49	46
DeepSeek V4 Pro	65	67	38	57		69	92	59	67	73
GLM-5.3	76	52	36	44	69		73	50	67	86
Kimi K3	64	68	38	51	92	73		54	72	70
MiniMax M3	46	40	34	73	59	50	54		53	56
Mistral Large 4	57	42	38	49	67	67	72	53		67
Nemotron 3 Ultra	79	49	38	46	73	86	70	56	67