Operator · San Francisco, CA
Replicate
US model-hosting platform, joining Cloudflare since November 2025, that runs public and custom Cog-packaged models behind an API, billed per second of GPU hardware time or per output for some models.
GPU types5with a current offer
Current offers3738 readings on file · last 45 days
H100 from$5.49per GPU-hour, on-demand
Entry price$0.81cheapest GPU-hour listed
Countries1regions not published
Trust Score v0.169/100
- Company & track record73
- Infrastructure control100
- Pricing & billing transparency65
- Reliability signals85
- Security & compliance20
- Customer signal—
Scored Sep 29, 2026 from the provider's own pages. Dimensions not yet measured are excluded from the denominator. Criteria ↓
Trust Score · 5 of 6 dimensions measuredThe score, criterion by criterion
Each point comes from something the provider publishes, with the page it was read on and the date. Anything not disclosed earns zero.
Company & track record
weight 15%
73 / 100
| Founding year published | not disclosedreplicate.com · Sep 29, 2026 | 0 / 4 |
| Headquarters published | Replicate, LLCreplicate.com · Sep 29, 2026 | 3 / 3 |
| Legal entity named | Replicate, LLCreplicate.com · Sep 29, 2026 | 4 / 4 |
| Funding or ownership disclosed | Joining Cloudflare announced 17 Nov 2025; Replicate continues as a distinct brand. Careers listed on the Cloudflare careers page.replicate.com · Sep 29, 2026 | 4 / 4 |
Infrastructure control
weight 15%
100 / 100
| Capacity model published | operates or rents dedicated capacityreplicate.com · Sep 29, 2026 | 10 / 10 |
| Regions published | 1 regions or countriesreplicate.com · Sep 29, 2026 | 5 / 5 |
Pricing & billing transparency
weight 20%
65 / 100
| Public prices without an account | yes (read on the pricing page)replicate.com · Sep 29, 2026 | 5 / 5 |
| Billing granularity published | per secondreplicate.com · Sep 29, 2026 | 4 / 4 |
| Egress policy published | not disclosedreplicate.com · Sep 29, 2026 | 0 / 4 |
| Stopped-instance billing published | No VM to stop. Models scale to zero (min instances 0 by default) and nothing is billed while offline. Public models: only active processing time is billed. Private models and deployments: setup, idle (instances stay idle a few minutes after the last request) and active time are billed; a deployment with min_instances above 0 is billed continuously.replicate.com · Sep 29, 2026 | 4 / 4 |
| Public IPv4 policy published | not published (API endpoints only, no VM or IP address)replicate.com · Sep 29, 2026 | 0 / 3 |
Reliability signals
weight 20%
85 / 100
| Status page | https://www.replicatestatus.com/replicate.com · Sep 29, 2026 | 6 / 6 |
| SLA published | Uptime and performance SLAs offered to enterprise customers; levels not publishedreplicate.com · Sep 29, 2026 | 6 / 6 |
| SLA with a numeric uptime commitment | not disclosedreplicate.com · Sep 29, 2026 | 0 / 3 |
| Uptime history published | Cloudflare Status shows Replicate operational, 5 days with incidents in the last 90 days (29 Sep 2026); no uptime percentage.replicate.com · Sep 29, 2026 | 5 / 5 |
| Observed availability (30 days) | not measured yet | — |
Security & compliance
weight 15%
20 / 100
| SOC 2 | not disclosedreplicate.com · Sep 29, 2026 | 0 / 8 |
| ISO 27001 | not disclosedreplicate.com · Sep 29, 2026 | 0 / 4 |
| HIPAA or GDPR DPA | Data processing agreements offered to enterprise customers; subprocessor list published.replicate.com · Sep 29, 2026 | 3 / 3 |
Customer signal
weight 15%
not measured
| Aggregated public reviews | not collected yet | — |
provisional v0.1: evidence from the provider's own pages; dimensions not yet measured are excluded from the denominatorFull method →
Prices by GPU · 5 GPU types with a current offerWhat Replicate charges, against the market
Cheapest current on-demand true cost here, next to the market median for the same GPU. Negative means cheaper than the market.
| GPU | Cheapest here | Market median | vs market | Current offers | Modes | |
|---|---|---|---|---|---|---|
| H100 | $5.49 | $3.17 | +73% | 4 | on-demand | H100 prices |
| A100 80GB | $5.04 | $1.99 | +153% | 4 | on-demand | A100 80GB prices |
| H200 | $5.49 | $4.76 | +15% | 4 | on-demand | H200 prices |
| L40S | $3.51 | $1.26 | +178% | 4 | on-demand | L40S prices |
| T4 | $0.81 | $0.51 | +59% | 1 | on-demand | T4 prices |
All offers · 38 readings on fileEvery Replicate offer, with its receipt
Current offers (last 45 days) ranked by true cost, all rental modes, each with its read date. Older readings stay listed below, in grey.
| GPU | Configuration | Mode | Region | Stock | List | True cost | Read | |
|---|---|---|---|---|---|---|---|---|
| T4 | 1× T4 | On-demand | — | Stock unknown | $0.81 | $0.81 | 21:35 UTCfresh | Receipt |
| L40S | 1× L40S | On-demand | — | Stock unknown | $3.51 | $3.51 | 21:35 UTCfresh | Receipt |
| L40S | 2× L40S | On-demand | — | Stock unknown | $3.51 | $3.51 | 21:35 UTCfresh | Receipt |
| L40S | 4× L40S | On-demand | — | Stock unknown | $3.51 | $3.51 | 21:35 UTCfresh | Receipt |
| L40S | 8× L40S | On-demand | — | Stock unknown | $3.51 | $3.51 | 21:35 UTCfresh | Receipt |
| A100 80GB | 1× A100 80GB | On-demand | — | Stock unknown | $5.04 | $5.04 | 21:35 UTCfresh | Receipt |
| A100 80GB | 2× A100 80GB | On-demand | — | Stock unknown | $5.04 | $5.04 | 21:35 UTCfresh | Receipt |
| A100 80GB | 4× A100 80GB | On-demand | — | Stock unknown | $5.04 | $5.04 | 21:35 UTCfresh | Receipt |
| A100 80GB | 8× A100 80GB | On-demand | — | Stock unknown | $5.04 | $5.04 | 21:35 UTCfresh | Receipt |
| H100SXM | 1× H100 SXM | On-demand | — | Stock unknown | $5.49 | $5.49 | 21:35 UTCfresh | Receipt |
| H100SXM | 2× H100 SXM | On-demand | — | Stock unknown | $5.49 | $5.49 | 21:35 UTCfresh | Receipt |
| H100SXM | 4× H100 SXM | On-demand | — | Stock unknown | $5.49 | $5.49 | 21:35 UTCfresh | Receipt |
| H100SXM | 8× H100 SXM | On-demand | — | Stock unknown | $5.49 | $5.49 | 21:35 UTCfresh | Receipt |
| H200 | 1× H200 | On-demand | — | Stock unknown | $5.49 | $5.49 | 21:35 UTCfresh | Receipt |
| H200 | 2× H200 | On-demand | — | Stock unknown | $5.49 | $5.49 | 21:35 UTCfresh | Receipt |
| H200 | 4× H200 | On-demand | — | Stock unknown | $5.49 | $5.49 | 21:35 UTCfresh | Receipt |
| H200 | 8× H200 | On-demand | — | Stock unknown | $5.49 | $5.49 | 21:35 UTCfresh | Receipt |
17 fresh offers · prices per GPU-hour · true cost includes the add-ons the provider publishesOpen in the comparison tool →
True cost & billing · read Sep 29, 2026What the hourly price does not say
Billing granularity, minimums, what a stopped machine costs, traffic, addresses, deposits and identity checks, as published by the provider.
Billing
- Billing granularity
- per second
- Minimum charge
- not published
- Stopped instance
- No VM to stop. Models scale to zero (min instances 0 by default) and nothing is billed while offline. Public models: only active processing time is billed. Private models and deployments: setup, idle (instances stay idle a few minutes after the last request) and active time are billed; a deployment with min_instances above 0 is billed continuously.
- Egress
- not published
- Public IPv4
- not published (API endpoints only, no VM or IP address)
- Billing model
- both
- Minimum deposit
- Not published. Auto reload minimum threshold 5 USD, minimum reload balance 15 USD. Purchased credit valid 1 year from purchase, non-refundable.
- Free credits
- Select models can be run for free for a while, then billing must be set up (amount not published).
Account
- Payment
- card · invoice · crypto (not published)
- Sign-up
- GitHub account (the sign-in page offers only 'Sign in with GitHub'), age 18 or over (terms), payment method (credit card, debit card or bank transfer) to set up billing after free limits
- Identity check
- not published
- Support
- email ([email protected]), support page (sign-in required), Discord community
- Enterprise support
- not published
- SLA
- Uptime and performance SLAs offered to enterprise customers; levels not published
- Status page
- replicatestatus.com ↗
- Published uptime
- Cloudflare Status shows Replicate operational, 5 days with incidents in the last 90 days (29 Sep 2026); no uptime percentage.
Infrastructure
- Regions
- United States
- Countries
- United States
- Platform
- API, CLI, serverless, custom Docker imagesSDK: python, node.js, go, swift (official); MCP server; community elixir, ruby, rails, laravelKubernetes: noClusters: No cluster product; deployments autoscale from zero to hundreds of instances. 4x/8x A100, 2x to 8x H100, all H200 and 4x/8x L40S need committed spend contracts.
- Storage
- not published
- Migration out
- Models are packaged as standard Docker containers with the open-source Cog tool, so the same image can run on any Docker GPU host. API prediction outputs are deleted after one hour, so customers already keep their own copies. There are no volumes or disks to export. Moving off means replacing Replicate HTTP API calls and its autoscaling with a self-managed inference server.
Security & compliance
- SOC 2
- not published
- ISO 27001
- not published
- HIPAA
- not published
- GDPR / DPA
- Data processing agreements offered to enterprise customers; subprocessor list published.
- Other
- API prediction inputs, outputs and logs deleted automatically after one hour; web predictions kept indefinitely. Indemnity coverage available in enterprise contracts.
- Legal entity
- Replicate, LLC
- Funding
- Joining Cloudflare announced 17 Nov 2025; Replicate continues as a distinct brand. Careers listed on the Cloudflare careers page.
Read on the provider's own pages · 21 pages read · facts marked "not published" were looked for and not foundPages read ↓
Assessment · documents read Sep 29, 2026Replicate, in short
What stood out when we read the documents: strengths, weaknesses, who it suits, and what we could not find.
Strengths
- Per-second billing with scale-to-zero: public models bill only active processing time, setup and idle time are free.
- Large model catalogue: the enterprise page cites 50,000+ models behind one API, including proprietary models (OpenAI, Google, Black Forest Labs, Anthropic) sold per output.
- Open-source Cog packaging (standard Docker containers), official Python, Node.js, Go and Swift clients, an official CLI and an MCP server.
- Deployments with autoscaling, min/max instances, rolling updates and hardware switching via API.
- Owned by Cloudflare since the November 2025 announcement.
Weaknesses
- High per GPU-hour rates: H100 5.49 USD, A100 80GB 5.04 USD, L40S 3.51 USD, T4 0.81 USD.
- No raw GPU access: no VM, SSH, Jupyter or persistent disk; workloads must be packaged as Cog models.
- 4x/8x A100 and L40S, every multi-GPU H100 and all H200 SKUs need committed spend contracts.
- Private models and deployments bill setup and idle time.
- Prepaid credit is non-refundable and expires after 12 months; no crypto payment published.
- No regions published; sign-in only via GitHub; no SOC 2 or ISO 27001 report found on public pages.
Ideal for
- Developers who want to call hosted open-source or proprietary models through an API without managing GPUs.
- Bursty or low-volume inference where scale-to-zero outweighs a higher hourly rate.
- Teams publishing custom models packaged with Cog.
Not published
- H100 form factor (SXM, PCIe or NVL) not published
- A100 80GB form factor (SXM or PCIe) not published
- H200 GPU memory and CPU/RAM for the 'Additional hardware' SKUs not listed.
- Data center regions not published
- Crypto payment, PayPal, minimum first credit purchase and free credit amount not published.
- No SOC 2, ISO 27001 or HIPAA statement found
- trust.replicate.com does not resolve and replicate.com/security returns 404.
- Enterprise SLA levels, volume discounts and committed-spend contract terms not published.
- Egress or bandwidth charges not published.
- Founding year not published on the pages read.
Observations
Pricing page is public; no account needed to see prices.
The page opens with per-output prices for example models (e.g. flux-1.1-pro 0.04 USD per output image, deepseek-r1 3.75 USD per million input tokens) before the hardware table.
Hardware table lists each SKU with price per second and per hour, GPU count, CPU count, GPU RAM and RAM; CPU-only SKUs (cpu-small 0.09 USD/hr, cpu 0.36 USD/hr) included.
An 'Additional hardware' block lists 4x/8x A100, 2x/4x/8x H100, 1x to 8x H200 and 4x/8x L40S with prices but no specs, each marked 'available with committed spend contracts'.
No region selector, and no storage, egress or IP address lines.
Top banner promotion 'Wan 3.0 is 30% off this week'; primary CTA 'Try for free'.
Enterprise volume discounts mentioned without figures; footer shows 'All services are online'.
Pages read
- replicate.com/pricing — Hardware per-second prices, SKUs, specs, per-output model prices, committed-spend gating
- replicate.com/ — Positioning, scale-to-zero, pricing excerpt
- replicate.com/docs/topics/billing — Billing models, instance lifecycle, what is billed for public/private models and deployments, failed runs, free limits
- replicate.com/docs/topics/billing/prepaid-credit — Prepaid credit, 1-year validity, non-refundable, auto reload minimums
- replicate.com/terms — Legal entity and address, payment methods, Stripe, arrears vs prepaid, 18+ eligibility, export controls
- replicate.com/privacy — GitHub sign-in data, legal entity
- replicate.com/signin — Sign-in method (GitHub only
- replicate.com/enterprise — SLA, DPA, priority support, volume discounts, reserved compute
- replicate.com/blog/replicate-cloudflare — Ownership
- replicate.com/about — Careers on Cloudflare
- replicate.com/docs/topics/deployments — Autoscaling, hardware choice, monitoring
- replicate.com/docs/topics/deployments/create-a-deployment — Deployment API, min/max instances
- replicate.com/docs/topics/models/hardware — Hardware switching
- replicate.com/docs/reference/client-libraries — SDKs, MCP
- replicate.com/docs/reference/open-source — Cog, official CLI
- replicate.com/docs/reference/http — API endpoints incl. hardware.list
- replicate.com/docs/reference/how-does-replicate-work — Cold boots
- replicate.com/docs/topics/predictions/data-retention — One-hour deletion of API prediction data
- replicate.com/docs/topics/predictions/rate-limits — 600 prediction requests per minute, credit-based throttling
- replicate.com/docs/topics/site-policy/subprocessors — Infrastructure providers and locations
- replicatestatus.com/ — Status page, 90-day incident history
Questions about Replicate
What GPUs does Replicate offer?
Replicate has 5 GPU types with a fresh offer today, for example H100 from $5.49, A100 80GB from $5.04, H200 from $5.49, L40S from $3.51, T4 from $0.81 per GPU-hour on-demand. The full list with prices and market medians is on this page.
How does Replicate bill?
Billing granularity: per second. Stopped instance: No VM to stop. Models scale to zero (min instances 0 by default) and nothing is billed while offline. Public models: only active processing time is billed. Private models and deployments: setup, idle (instances stay idle a few minutes after the last request) and active time are billed; a deployment with min_instances above 0 is billed continuously.
How trustworthy is Replicate?
Its Trust Score is 69/100 (v0.1), read from its own documents. Strong on infrastructure control. Weaker on pricing & billing transparency, security & compliance. The score measures transparency and published guarantees, not customer satisfaction; every criterion and its source is listed above.
Where are Replicate's data centres?
1 countries: United States.
How do you pay Replicate, and are there credits?
Payment by card, invoice, crypto. Free credits: Select models can be run for free for a while, then billing must be set up (amount not published)..