Replicate review

Replicate is the broadest catalog in this list but the weakest fit for low-latency, per-token LLM serving. Thousands of community-contributed and proprietary models run through one API, and most are billed by the second of hardware time, from $0.000225 per second on an Nvidia T4 to $0.001525 on an H100. Some models are billed per input and output instead, and each model page shows a cost estimate. You can deploy your own models with Cog, Replicate's open-source packaging tool, and deployments scale up and down automatically. The costs to watch are on private models: most run on dedicated hardware and bill for setup and idle time as well as active time, except fast booting fine-tunes. Token prices on popular LLMs, such as deepseek-r1 at $3.75 per 1M input tokens, are well above the specialists, which is why Replicate ranks last for this audience.

Who it is for

Developers running image, video and niche community models

Price

Pay for what you use; public models billed per second of hardware or per input and output; Nvidia T4 at $0.000225 per second; deepseek-r1 at $3.75 per 1M input tokens.

Nvidia T4 GPU gpu-t4 $ 0.000225 /sec $ 0.81 /hr
Quoted from the vendor pricing page on November 12, 2026.

Key features

• Thousands of open-source models contributed by the community, plus hosted proprietary models
• Cog, an open-source tool for packaging and deploying your own models
• Hardware from CPU to 8x H100 and 8x H200 billed per second
• Fast booting fine-tunes billed only while processing requests
• Automatic scaling up and down for deployed models

Strengths and weaknesses

• Strength: Very large catalog of image, video and audio models
• Strength: Per-model cost estimates shown on each model page
• Strength: Nvidia H100 at $5.49 per hour, billed per second
• Weakness: Private models are billed for setup and idle time as well as active time
• Weakness: Per-token LLM pricing is less competitive, for example deepseek-r1 at $3.75 per 1M input tokens
• Weakness: Multi-GPU A100, H100 and H200 capacity beyond 1x often requires committed spend

Key facts

• Nvidia T4: $0.000225 per second, $0.81 per hour
• Nvidia H100: $0.001525 per second, $5.49 per hour
• Custom models: Packaged with Cog
• Billing: Hardware time or per input and output, depending on the model

Where it ranks

Replicate is No. 10 in our Best AI Inference Platforms verdict.