Best LLM Inference Providers

Vendor profileChecked Nov 8, 2026

Cloudflare Workers AI

Cloudflare, Inc. (Cloudflare also owns Replicate, profiled separately) · developers.cloudflare.com · Last checked:

  • Serverless per-token API on Cloudflare's network

Company status

Active, publicly listed on the NYSE (NET). Workers AI is a Cloudflare product.

Sources 1, 2, 3, 4, 5, 6, 7, 8 · checked 2026-11-08

Capabilities

  • Catalog of 50+ open-source (open-weight) models served on GPUs across Cloudflare's network, callable from Workers or the REST API
  • OpenAI-compatible endpoints for Chat Completions and embeddings; Responses API only for GPT-OSS models
  • Fine-tuned inference with uploaded LoRA adapters (open beta; Mistral, Gemma, and Llama base models)
  • Asynchronous Batch API (beta)
  • Zero data retention for standard inference requests; Cloudflare states it does not use Workers AI customer content to train models

Sources 1, 2, 3, 4, 5, 6, 7, 8 · checked 2026-11-08

Who it tends to fit

  • Teams already building on Cloudflare Workers that want inference in the same platform
  • Small or spiky workloads that fit a daily free allocation plus pay-per-use

Sources 1, 2, 3, 4, 5, 6, 7, 8 · checked 2026-11-08

Pricing

$0.011 per 1,000 Neurons beyond a free allocation of 10,000 Neurons per day; usage above the free allocation requires the Workers Paid plan. Per-model token prices are listed (for example @cf/meta/llama-3.1-8b-instruct-fp8-fast at $0.045 per M input and $0.384 per M output tokens).

Sources 1, 2, 3, 4, 5, 6, 7, 8 · checked 2026-11-08

Published certifications

  • SOC 2 Type II
  • ISO 27001:2022
  • FedRAMP High
  • PCI DSS 4.0
  • HIPAA (AoC on request)
  • GDPR (EU Cloud Code of Conduct)
  • C5:2020

Listed on Cloudflare's company-wide compliance page, which does not say which products each covers; confirm Workers AI is in scope.

Listed as stated by the vendor. Scope can differ by product or subsidiary; request current reports.

Sources 1, 2, 3, 4, 5, 6, 7, 8 · checked 2026-11-08

Limitations to check

  • LoRA adapters and the Batch API are in beta
  • Catalog is limited to models Cloudflare hosts; LoRA adapters must target supported base models
  • Stateful features such as AI Gateway logging or caching store content under their own retention rules

Sources 1, 2, 3, 4, 5, 6, 7, 8 · checked 2026-11-08

Sources

  1. Overview: serverless GPUs on Cloudflare's network, 50+ open-source models, Free and Paid plansdevelopers.cloudflare.com/workers-ai/Checked 2026-11-08
  2. Pricing: $0.011 per 1,000 Neurons, 10,000 Neurons/day free, Workers Paid needed above that; per-model token pricesdevelopers.cloudflare.com/workers-ai/platform/pricing/Checked 2026-11-08
  3. OpenAI-compatible Chat Completions and embeddings; Responses API for GPT-OSS onlydevelopers.cloudflare.com/workers-ai/configuration/open-ai-compatibility/Checked 2026-11-08
  4. LoRA adapters: open beta, model_type mistral, gemma or llamadevelopers.cloudflare.com/workers-ai/features/fine-tunes/loras/Checked 2026-11-08
  5. Asynchronous Batch API (beta)developers.cloudflare.com/workers-ai/features/batch-api/Checked 2026-11-08
  6. Data usage: zero data retention for standard requests; no training on customer content; exceptions for stateful featuresdevelopers.cloudflare.com/workers-ai/platform/data-usage/Checked 2026-11-08
  7. Compliance page: ISO 27001:2022, FedRAMP High, SOC 2 Type II, PCI DSS 4.0, HIPAA, C5:2020, EU Cloud Code of Conduct (GDPR)www.cloudflare.com/trust-hub/compliance-resources/Checked 2026-11-08
  8. Press release identifies Cloudflare, Inc. as NYSE: NETwww.cloudflare.com/press/press-releases/2025/cloudflare-to-acquire-replicate-to-build-the-most-seamless-ai-cloud-for-developers/Checked 2026-11-08

Report a correction: [email protected]. Facts are re-checked at least every 90 days.