Pricing & SLA
Chat API pricing & SLA, side by side
What you get for free, what the first paid plan costs and what each vendor promises on uptime. Every competitor figure comes from its own pricing page.
// 01 — TABLE
Free tier, first paid plan and uptime SLA
| Service | Free tier: MAU | Free tier: peak concurrent connections | First paid plan (monthly billing; annual in parentheses) | Uptime SLA |
|---|---|---|---|---|
| GravEx Chat (Basic) | No cap enforced through pricing | No cap enforced through pricing. Each app has a safety limit of 10 concurrent connections per user. | No paid plan. Basic is $0 with no card, and includes every engine feature. Price changes are announced at least 90 days ahead. | 99.5% monthly uptime target. It is an operating target with no credits or compensation if missed, not a credit-backed guarantee. |
| Sendbird | 1,000 (Developer plan). No overage. | 20 | Starter: $399/mo ($349/mo billed annually). 5,000 MAU; concurrent connections capped at 5% of MAU. Pro: $599/mo ($499/mo billed annually). | No uptime SLA percentage is listed on the plans. |
| Stream | 1,000 | 100 | Start: $499/mo ($399/mo billed annually). 10,000 MAU and 500 concurrent connections. | 99.999% SLA only on Enterprise or as a paid add-on. |
| CometChat | 100 (Build plan, for development and testing) | 25 | Basic: $239/mo, billed annually. Concurrent connections capped at 5% of MAU; overage is $0.10 per MAU. | 99.999% uptime only as an add-on on the Advanced plan. |
| PubNub | 200 MAU or 1M transactions (for development and testing) | Not stated on that pricing page | Starter: $98/mo, including 1,000 MAU. | An uptime SLA (up to 99.999%) only on the Pro plan. |
Per each vendor's official pricing page, checked 2026-10-31. Vendors change plans and prices, so confirm on the linked page before you decide.
// 02 — NOTES
How to read this table
- MAU, concurrent connections and transactions are defined differently by each vendor, so the columns are a guide, not an exact like-for-like comparison.
- The table compares the free tier and the first paid plan only. Higher tiers, add-ons and enterprise contracts are priced by each vendor and are not shown.
- GravEx Chat does not sell an uptime guarantee. If you need a contractual SLA with credits, or compliance certifications, the enterprise plans of the larger vendors are the better fit.
// 03 — PERFORMANCE
What we measured
These are load-test results of the GravEx Chat engine. They were not measured under the same conditions as the vendors above, so they are not in the table. How we measured and what we did not measure are written up in Architecture and benchmarks.
| Setup | Result |
|---|---|
| One node limited to 0.2 vCPU on an AWS EC2 t3.medium; one room, 300 connections, 30 publishers | About 5,900 frames/s delivered at 20 publishes/s. Three such nodes: about 18,000 frames/s, exactly 3.0x. |
| Many rooms, 2 connections each; PostgreSQL limited to 0.3 cores, three nodes | About 390 publishes/s sustained. PostgreSQL is the limit; roughly 1,300 publishes/s per PostgreSQL core (a figure extrapolated from the runs above, not measured at one full core). |
| One node, one room with 400 connections, on one development machine (Apple M5 Pro, 18 cores) that also ran PostgreSQL, Redis and the load generator | About 158,000 frames/s at 400 publishes/s; p99 delivery 52 ms, 0 dropped connections. A lower bound, not a production figure. |
// 04 — WHAT YOU GET
What you get
- Free: Basic is $0 with no card and no trial clock.
- Documentation and this site in Korean and English.
- Start with AI: your AI tool reads llms.txt and the console API, creates the app, issues the keys and wires up chat.
- Take your data with you: one click, or GET /v1/export, downloads every user, room and message as NDJSON.
- Any paid plan or price change is announced at least 90 days before it applies; a shutdown at least 180 days ahead, with export available for the whole period.