OpenRouter is the LLM gateway with the broadest model catalog (300+ models from Anthropic, OpenAI, Google, Mistral, and more) and a 5% routing markup over each model's native price. The OpenAI-compatible API and auto-fallback between providers are strong differentiators. Where alternatives win: LiteLLM is free OSS for self-hosting, Portkey bundles prompt management plus observability plus guardrails, Cloudflare AI Gateway is free up to 10K requests per day for Workers users, TrueFoundry is the full LLM-Ops platform, Vellum is workflow-first with evals, and Lunary is OSS-friendly with self-host option.
By Subrupt EditorialPublished Reviewed
LLM gateways emerged as a category in 2023 when teams realized that hardcoding a single model provider into apps creates fragility (rate limits, outages, model deprecation) and cost lock-in (no negotiation power, no provider arbitrage). The pattern: route LLM calls through a gateway that handles caching, fallback, observability, and cost tracking, while letting the underlying provider mix change without app code changes. OpenRouter took the marketplace approach; LiteLLM took the OSS proxy approach; Portkey and Vellum took the platform approach.
Pricing math: a typical app routing 10M tokens monthly through GPT-4o ($2.50 per 1M tokens input plus $10 per 1M tokens output) runs roughly $125-$250 in raw model cost. OpenRouter adds 5% on top ($6-$12 markup). Portkey Production at $49 monthly + 100K requests included covers most app traffic at fixed cost. LiteLLM OSS is free if self-hosted (you pay only for the proxy infrastructure plus model API costs). The right choice depends less on absolute markup and more on which platform features your team needs.
Pick by your shape. Self-hosted Python proxy with full control: LiteLLM. Prompt management plus observability plus guardrails: Portkey. Cloudflare Workers-native at zero cost: Cloudflare AI Gateway. ML platform plus LLM-Ops bundled: TrueFoundry. Workflow builder with evals: Vellum. OSS observability with optional self-host: Lunary.
Affiliate disclosure: Subrupt earns a commission when you switch to a service through our recommendation links. This never changes the price you pay. We only recommend services where there's a real cost or feature advantage for you, and our picks are based on the data on this page, not on which programs pay the most.
Quick pick by use case
If you only have thirty seconds, find your situation below and skip to that pick.
LiteLLM is MIT-licensed Python proxy that turns 100+ LLM provider APIs into one OpenAI-compatible endpoint. Self-hosted is fully free (you pay only for the proxy infrastructure). Cloud Free tier is hosted with limits; Cloud Pro at $50 per user monthly adds cost tracking, budgets, virtual keys, and team management. For teams with strong DevOps capacity who want full control over the proxy and zero markup, LiteLLM OSS is the lowest absolute cost. The trade vs OpenRouter: you operate the proxy (Postgres for state, Redis for caching), the dashboard is functional rather than polished, and the routing intelligence is shallower than OpenRouter's auto-fallback.
Strengths
+MIT-licensed self-hostable
+100+ provider integrations
+Cost tracking + budgets on Cloud Pro
+Virtual API keys for team management
Trade-offs
−Operational overhead for self-hosting
−Less polished dashboard than OpenRouter
−Routing intelligence shallower than OpenRouter
OSS
Free, self-hosted MIT
Cloud Free
Limited request volume
Cloud Pro
$50/user/mo
Enterprise
Custom + SOC 2
Migration steps
Self-host LiteLLM proxy (Docker + Postgres + Redis) or sign up for Cloud.
Configure provider API keys in LiteLLM config.yaml.
Switch app's OpenAI base URL to LiteLLM proxy URL.
Validate cost tracking and virtual keys for team usage.
Cancel OpenRouter when LiteLLM covers routing needs.
Not for: LiteLLM is the wrong fit for teams without DevOps capacity to operate a proxy or those who need OpenRouter's marketplace breadth (300+ models vs LiteLLM's 100+); OpenRouter fits those better.
Portkey bundles LLM gateway, prompt management, observability, guardrails, and RBAC in one platform. Developer Free covers 10K requests monthly; Production at $49 monthly covers 100K requests; Pro at $249 covers 1M requests plus guardrails plus SSO; Enterprise adds on-prem self-hosted. The differentiator is the bundled platform: where OpenRouter is gateway-only, Portkey ships with the dashboard, prompt versioning, evals, and observability that most teams build separately. The trade vs OpenRouter: smaller model catalog (40+ vs 300+), Production tier surprises high-volume teams above 100K requests.
Strengths
+Bundled gateway + prompts + observability
+10K free requests + 40+ providers
+Guardrails on Pro tier
+SOC 2 Type II + HIPAA on Enterprise
Trade-offs
−Smaller model catalog than OpenRouter
−Production tier 100K request limit can compound
−Dashboard learning curve
Free
10K requests/mo
Production
$49/mo, 100K requests
Pro
$249/mo, 1M requests + guardrails
Enterprise
Custom + on-prem
Migration steps
Sign up at portkey.ai (free).
Configure Virtual Keys for each provider.
Switch app's API endpoint to Portkey gateway.
Set up prompt registry and observability dashboards.
Cancel OpenRouter if Portkey's 40+ providers cover your model needs.
Not for: Portkey is the wrong fit for teams who need OpenRouter's 300+ model marketplace breadth or those who only want gateway routing without bundled platform features.
Cloudflare AI Gateway is free up to 10K requests per day with caching, analytics, and rate limits. Workers Paid at $5 monthly raises request limits; Workers Standard at $200 monthly includes 50M Workers requests with unlimited AI Gateway. The platform is purpose-built for apps running on Cloudflare Workers (and works with non-Workers apps too). For Cloudflare-native teams whose stack already pays for Workers, the AI Gateway is essentially free incremental capability. The trade vs OpenRouter: less polished prompt management UI, smaller model catalog (focused on OpenAI, Anthropic, Google, Workers AI), and the platform assumes Cloudflare-friendly stack.
Strengths
+Free up to 10K requests/day
+Caching + analytics + rate limits built in
+Native Workers integration
+Bundled with Cloudflare Workers paid tier
Trade-offs
−Less polished prompt management than Portkey
−Smaller model catalog than OpenRouter
−Best fit only for Cloudflare-friendly stacks
Free
10K requests/day
Workers Paid
$5/mo + higher limits
Workers Standard
$200/mo + unlimited AI Gateway
Enterprise
Custom + dedicated regions
Migration steps
Sign up for Cloudflare account (free tier).
Create AI Gateway in Cloudflare dashboard.
Switch app's OpenAI/Anthropic base URL to Cloudflare AI Gateway URL.
Configure caching and rate limit rules.
Cancel OpenRouter when Cloudflare covers your routing needs.
Not for: Cloudflare AI Gateway is the wrong fit for teams who need OpenRouter's full marketplace catalog or who do not run on Cloudflare; OpenRouter or Portkey cover those better.
TrueFoundry is the full LLM-Ops platform with gateway, prompt management, observability, evals, and self-hosted enterprise tier. Cloud Pro at $300 monthly + usage covers prompt management, evals, custom dashboards, and Slack alerts. Self-Hosted at $1.5K monthly runs on your Kubernetes with SSO and audit logs. Enterprise at $4K monthly covers multi-region with dedicated CSM. For ML platform teams who want one vendor relationship covering both LLM-Ops and traditional ML model serving, TrueFoundry consolidates the stack. The trade vs OpenRouter: 10x the cost, 10-20x more platform features (often more than small teams need), longer onboarding.
Strengths
+Full LLM-Ops + ML platform bundled
+Self-hosted enterprise tier
+Custom dashboards + Slack alerts
+SOC 2 + HIPAA on Enterprise
Trade-offs
−10x cost vs OpenRouter base
−Onboarding takes 2-4 weeks for enterprise
−Overkill for gateway-only needs
Free
Trial credits
Cloud Pro
$300/mo + usage
Self-Hosted
$1.5K/mo on your K8s
Enterprise
Custom + dedicated CSM
Migration steps
Schedule sales call with TrueFoundry (1-2 weeks).
Configure gateway and prompt registry.
Migrate observability dashboards.
Run parallel with OpenRouter for 30 days.
Cancel OpenRouter and any standalone observability tools after parity.
Not for: TrueFoundry is the wrong fit for teams who only want gateway routing without the broader LLM-Ops platform; OpenRouter, Portkey, or LiteLLM cover those better.
Vellum is the workflow-first LLM platform with prompt playground, visual workflow builder, evals, and API deployments. Pro at $500 monthly covers workflows + evals + deployments; Business at $2K covers RBAC + SSO + audit logs; Enterprise covers on-prem deploy. The differentiator vs OpenRouter is the workflow surface: where OpenRouter is gateway-only, Vellum lets you build visual chains (prompt -> tool call -> prompt -> output) with evals attached. The trade vs OpenRouter: 25x the cost, smaller model catalog, and the workflow surface is a learning curve for teams who prefer code-first orchestration.
Strengths
+Visual workflow builder
+Built-in evals + deployment
+API deployments with versioning
+Prompt playground for non-engineers
Trade-offs
−25x cost vs OpenRouter
−Workflow surface learning curve
−Smaller model catalog
Free
Trial credits
Pro
$500/mo + usage
Business
$2K/mo + SSO
Enterprise
Custom + on-prem
Migration steps
Sign up at vellum.ai (trial).
Build representative workflows in Vellum visual builder.
Migrate prompt registry from OpenRouter or other tools.
Deploy via Vellum API and switch app traffic.
Cancel OpenRouter when Vellum covers gateway plus workflow needs.
Not for: Vellum is the wrong fit for teams whose primary need is gateway routing without workflow building; OpenRouter, LiteLLM, or Portkey cover those better.
Paid plans from $500.00/mo
When to stay with OpenRouter
Stay with OpenRouter if your app routes across 300+ models with auto-fallback already configured, your virtual key tracking is wired into reporting, or your routing markup of 5% over base model price stays under your cost ceiling. The picks below address self-hosted Python proxy, prompt-management-bundled, Cloudflare Workers-native, ML-platform-bundled, full LLM-Ops platform with workflows, and OSS observability with self-host option.
LLM gateway alternatives split along three vectors: deployment model (managed cloud vs OSS self-hosted vs Cloudflare-native), feature scope (gateway-only vs gateway+observability vs full LLM-Ops platform), and pricing model (markup-on-tokens vs flat-monthly vs request-tier). Picks below address each combination.
Pricing pulled from each vendor's site on the review date. We score on cost-at-volume for representative LLM apps (10M tokens monthly across mixed providers), feature breadth (gateway, prompt management, observability, evals, guardrails), provider catalog depth, and operational lift to migrate. We weight against tools whose advertised free tier excludes essential features that quickly push teams to paid.
Update history1 update
Initial published version with 5 picks.
Frequently asked questions about OpenRouter alternatives
Why use a gateway at all instead of calling provider APIs directly?
Three reasons: (1) provider failover - when OpenAI has an outage, automatically retry on Anthropic; (2) cost optimization - route to cheaper provider when quality difference is acceptable; (3) observability - centralized logging, cost tracking, and prompt versioning across providers. The cost (markup or subscription) typically pays back when you have multi-provider usage above $1K monthly. Below that, direct API calls work fine.
How do I evaluate LLM gateway markup vs subscription pricing?
OpenRouter charges 5% markup on token costs. Portkey Production at $49 covers 100K requests with no markup. Cloudflare AI Gateway is free up to 10K daily requests. Math: if you do 1M monthly requests at $0.05 each ($50K), OpenRouter adds $2.5K markup; Portkey Pro at $249 covers 1M requests included. The crossover is roughly $5K-$10K monthly traffic where flat-fee subscription beats markup.
Should I self-host LiteLLM or use a managed gateway?
Self-hosting works for teams with DevOps capacity and >$5K monthly LLM cost where the markup or subscription savings exceed operational overhead. Managed gateway works for teams under $5K monthly or those without DevOps capacity. The biggest hidden cost of self-hosting is observability infrastructure (Postgres for logs, Redis for caching) which managed gateways include. Most teams under 50 engineers find managed pays back vs DevOps time.
Can I use multiple gateways together?
Yes, sometimes. Pattern 1: OpenRouter for marketplace breadth + Portkey for observability (Portkey acts as middleware in front of OpenRouter). Pattern 2: Cloudflare AI Gateway for caching + LiteLLM for routing logic. Pattern 3: Direct provider APIs for high-volume routes + gateway for low-volume routes. Most teams pick one primary gateway and use direct provider APIs as escape hatch for specific routes.
What about prompt management as a separate tool from gateway?
Some teams pair gateway-only tools (OpenRouter, LiteLLM, Cloudflare) with separate prompt-management tools (Langfuse, PromptLayer, LangSmith). Other teams use bundled platforms (Portkey, Vellum, TrueFoundry) that include both. The bundled approach is simpler operationally; the separate approach gives best-of-breed depth. The choice depends on whether prompt management is a heavy daily-use surface (favor bundled) or an occasional review surface (favor separate).
SE
About the author: Subrupt Editorial
The team behind subrupt.com. We track subscriptions, surface cheaper alternatives, and publish comparisons where the score formula is on the page so you can recompute it yourself. We do not claim 30,000 hours of testing. What we claim is live pricing from our database, a transparent composite score, and honest savings math against a category baseline.
Get notified of price drops for OpenRouter
We'll email you when OpenRouter or its alternatives lower their prices.
Track OpenRouter and find more savings
Add OpenRouter to your dashboard to monitor spending and discover even more alternatives.