Replicate
replicate.com · AI · watching since Aug 17, 2026 · RSS
Current state
As of Aug 27, 2026Replicate is a pay-as-you-go platform with no fixed subscription plans; costs depend on model usage (per-second hardware time, or per input/output token/image) for public models, and dedicated hardware uptime for private models. Enterprise plans with volume discounts and dedicated support are available on request.
| Plan | Price | Includes |
|---|---|---|
| CPU (Small) | $0.000025 per second | $0.09/hr · 1x CPU, 2GB RAM |
| CPU | $0.000100 per second | $0.36/hr · 4x CPU, 8GB RAM |
| Nvidia T4 GPU | $0.000225 per second | $0.81/hr · 1x GPU, 4x CPU, 16GB GPU RAM, 16GB RAM |
| Nvidia L40S GPU | $0.000975 per second | $3.51/hr · 1x GPU, 10x CPU, 48GB GPU RAM, 65GB RAM |
| Nvidia A100 (80GB) GPU | $0.001400 per second | $5.04/hr · 1x GPU, 10x CPU, 80GB GPU RAM, 144GB RAM |
| Nvidia H100 GPU | $0.001525 per second | $5.49/hr · 1x GPU, 13x CPU, 80GB GPU RAM, 144GB RAM |
| Nvidia H200 GPU | $0.001525 per second | $5.49/hr · available with committed spend contracts |
| 2x Nvidia L40S GPU | $0.001950 per second | $7.02/hr · 2x GPU, 20x CPU, 96GB GPU RAM, 144GB RAM |
| 2x Nvidia A100 (80GB) GPU | $0.002800 per second | $10.08/hr · 2x GPU, 20x CPU, 160GB GPU RAM, 288GB RAM |
| 2x Nvidia H100 GPU | $0.003050 per second | $10.98/hr · available with committed spend contracts |
| 4x Nvidia L40S GPU | $0.003900 per second | $14.04/hr · available with committed spend contracts |
| Enterprise | — | Dedicated account manager · Priority support · Higher GPU limits · Performance SLAs · Onboarding & custom model help |
Changes
This is the Competitor pricing pack running on a real vendor.
Current state
As of Aug 25, 2026Replicate's changelog spans from March 2022 through April 2026, covering additions such as agent skills, MCP server improvements, prediction deadlines, and fallback models, alongside recurring 'The Little Things' weekly update rollups. Most recent entries focus on AI agent tooling (Agent Skills, MCP auto-discovery, Code mode) and API refinements (fallback models, source filtering, deadlines).
Deprecations
- Sunset 2025-07-01
Monthly spend limit setting for new customers
Latest entries
- Replicate publishes agent skills for coding assistantsaddition
- Nano Banana Pro adds fallback model support for rate limitsaddition
- MCP server supports auto-discovery via official MCP Registryaddition
- List predictions API can be filtered by sourceaddition
- The little things, week ending December 19, 2025change
- The little things, week ending December 5, 2025change
- The little things, week ending November 21, 2025change
- Code mode added for Replicate's MCP serveraddition
- The little things, week ending November 7, 2025change
- Deployment setup monitoring launched for enterprise customersaddition
- The little things, week ending October 24, 2025change
- Predictions can now be set with deadlinesaddition
Changes
This is the Product and API pack running on a real vendor.
Watch Replicate — or anything else — your way.
Your own prompt, cadence and channels. One alert with the before/after proof when your sentence comes true.