Is Replicate still available after the Cloudflare acquisition?

Yes. Cloudflare announced the acquisition on 2025-11-17 and closed it on 2025-12-01. replicate.com still serves models and lists per-image and per-second prices, for example $0.025 per FLUX.1 dev image and $0.000975 per second on an L40S (checked 2026-10-02).

What is the best Replicate alternative for running my own model code?

Modal or RunPod. Modal bills by the second (L40S $0.000542/s, about $1.95/hour) and runs your Python functions. RunPod rents pods by the hour (L40S from $0.79/hour on Community Cloud) and has a serverless option.

What is the best Replicate alternative for hosted image models?

fal.ai, if you want the fastest hosted inference and new closed models early (1,000+ models). Runware, if you want the lowest price per image (FLUX.1 dev $0.0038 at 1024x1024).

Which Replicate alternative has a flat monthly price?

ModelsLab. Its Open Source Unlimited plan is $149/month with unlimited generations on the open-source models it hosts and 15 parallel generations. Partner and closed models are billed separately, and there is no free plan.

When should I stay on Replicate?

When you need a very large catalogue (Cloudflare cites 50,000+ production-ready models) or you deploy custom models with Cog, Replicate's open-source packaging tool.

Comparisons · 2026-10-02 · ModelsLab Team

Best Replicate Alternatives 2026: 8 Image and Model APIs Compared

Replicate alternatives for image and model APIs: fal.ai, Modal, RunPod, Hugging Face, Together, DeepInfra, Runware and ModelsLab, with prices checked 2026-10.

Last updated 2026-10-02

Disclosure: stablediffusionapi.com is operated by the ModelsLab team. We list competitors where they are the better choice, and each price comes from the vendor's own page, linked in the tables.

Short answer (checked 2026-10-02): Replicate still works after Cloudflare bought it (deal closed 2025-12-01), and it is still the widest catalogue, with 50,000+ models by Cloudflare’s count. Teams that move usually go to fal.ai for faster hosted inference, Modal or RunPod to run their own code on GPUs (L40S from about $1.95/hour on Modal, $0.79/hour on RunPod Community Cloud), Runware for the lowest price per image, or ModelsLab for one flat $149/month price on open models.

Comparison table

Provider	Best for	Price (checked 2026-10)	Models	Notes
Replicate (baseline)	Model breadth, Cog custom deploys	FLUX.1 dev $0.025/image; L40S $0.000975/s (source)	50,000+ (Cloudflare figure)	Part of Cloudflare since 2025-12-01.
fal.ai	Speed, new closed models early	FLUX.1 dev $0.025/MP; H100 $4.50/hour list (source)	1,000+	Hosted models and GPU compute.
Modal	Your own Python code on GPUs	L40S $0.000542/s; H100 $0.001097/s (source)	Whatever you deploy	Starter plan includes $30 of compute a month.
RunPod	Cheapest raw GPU hours	L40S $0.79/hour Community, $1.09 Secure; RTX 4090 $0.34/hour Community (source)	Whatever you deploy	Pods and serverless endpoints.
Hugging Face Inference Endpoints	Deploying a Hub model as-is	L4 $0.80/hour; L40S $1.80/hour; H100 $4.50/hour (source)	Hub models	Dedicated endpoint per model.
Together AI	LLMs plus Flux on one key	FLUX.2 [dev] $0.0154/image (source)	Open LLMs, Flux, SDXL	Strongest on LLM inference.
DeepInfra	Cheap metered LLM and Flux calls	FLUX-1-dev $0.009 at 1024x1024, 25 steps (source)	Open LLMs, Flux	Priced by size and steps.
Runware	Low price per image	FLUX.1 dev $0.0038 at 1024x1024 (source)	400K+ in its registry	Image and video focus.
ModelsLab	Flat monthly price on open models, custom LoRAs	$21/month (3,250 calls), $47/month (10,000 calls), $149/month unlimited open models (source)	10,000+	Paid plans only. Partner models billed separately.

Best for each use case

You deploy custom models with Cog and need the long tail: stay on Replicate.
You want hosted models with low latency and new releases fast: fal.ai.
You want to run your own inference code: Modal (per-second, Python-native) or RunPod (cheapest GPU hours).
You want a Hub model behind a private URL: Hugging Face Inference Endpoints.
Your app is mostly LLM calls with some images: Together AI or DeepInfra.
You want the lowest metered price per image: Runware.
You generate a lot of images on open models and want a fixed bill: ModelsLab.

Provider notes

Replicate: when to stay

Replicate runs community and official models behind one API, and you deploy your own with Cog, its open-source packaging tool. Cloudflare announced the acquisition on 2025-11-17 and its 10-Q dates the close to 2025-12-01. Public models bill per output (FLUX.1 dev $0.025); private models bill for all the time their instances are online, so idle capacity costs money.

fal.ai

fal.ai lists 1,000+ production models and says its inference engine is “up to 10x faster” (vendor claim). For image models, its prices match Replicate’s for FLUX.1 and it often has new closed models first. It also rents GPUs (H100 $4.50/hour list on its pricing page).

Modal runs your Python functions on GPUs and bills per second: L40S $0.000542/s (about $1.95/hour), A100 80GB $0.000694/s, H100 $0.001097/s (pricing). It is the closest match to Replicate’s “push code, get an endpoint” model, without Cog.

RunPod

RunPod rents GPU pods by the hour (pricing): RTX 4090 from $0.34, L40S from $0.79, H100 SXM from $2.69 on Community Cloud. Its serverless product lists a 48 GB L40S-class worker at $1.75/hour, metered per second. Choose RunPod when GPU cost matters more than convenience.

Hugging Face Inference Endpoints

Inference Endpoints deploy a Hub model to a dedicated GPU: L4 $0.80/hour, L40S $1.80/hour, H100 $4.50/hour (pricing). Good for a model that is already on the Hub and needs a private URL.

Together AI and DeepInfra

Both are strongest on open LLM inference and also serve Flux. Together lists FLUX.2 [dev] at $0.0154 per image; DeepInfra bills FLUX-1-dev at $0.009 for a 1024x1024, 25-step image. Use them when one key should cover text and images.

Runware

Runware focuses on image and video inference at low prices: FLUX.1 dev is $0.0038 per 1024x1024 image, and its registry has 400K+ models. It is the price pick for metered image generation.

ModelsLab

ModelsLab (which runs this site) hosts 10,000+ models, including SD, SDXL, Flux and community checkpoints, on its own GPUs, and you can add custom LoRAs on the same API key. Plans: Basic $21/month (3,250 calls, 5 parallel), Standard $47/month (10,000 calls, 10 parallel), Open Source Unlimited $149/month (unlimited generations on the open models it hosts, 15 parallel). It wins when volume is high and the models are open. It is not a place to deploy your own code, and partner and closed models (for example Google or ByteDance models) are billed per call on top of the plan.

How we checked

We read each price from the vendor’s own pricing or model page on 2026-10-02. The Replicate model count comes from Cloudflare’s press release, because replicate.com does not publish a number. GPU hourly figures for Modal are the per-second price x 3,600.

FAQ

Is Replicate still available after the Cloudflare acquisition?: Yes. Cloudflare announced the acquisition on 2025-11-17 and closed it on 2025-12-01. replicate.com still serves models and lists per-image and per-second prices, for example $0.025 per FLUX.1 dev image and $0.000975 per second on an L40S (checked 2026-10-02).
What is the best Replicate alternative for running my own model code?: Modal or RunPod. Modal bills by the second (L40S $0.000542/s, about $1.95/hour) and runs your Python functions. RunPod rents pods by the hour (L40S from $0.79/hour on Community Cloud) and has a serverless option.
What is the best Replicate alternative for hosted image models?: fal.ai, if you want the fastest hosted inference and new closed models early (1,000+ models). Runware, if you want the lowest price per image (FLUX.1 dev $0.0038 at 1024x1024).
Which Replicate alternative has a flat monthly price?: ModelsLab. Its Open Source Unlimited plan is $149/month with unlimited generations on the open-source models it hosts and 15 parallel generations. Partner and closed models are billed separately, and there is no free plan.
When should I stay on Replicate?: When you need a very large catalogue (Cloudflare cites 50,000+ production-ready models) or you deploy custom models with Cog, Replicate's open-source packaging tool.