Alternatives hub · AI Gateway
LiteLLM alternatives
LiteLLM is free to self-host, but you run it yourself (PostgreSQL plus optional Redis) and pin its versions. The alternatives split into hosted gateways with a fee (OpenRouter 5.5%, Requesty 5%), free self-hosted gateways (Portkey, Bifrost, OmniRoute, 9Router) and Cloudflare AI Gateway's free layer in front of your own keys. Every row is checked against the vendor's page.
Our pick
Requesty or OpenRouter if you do not want to run a server, Bifrost or Portkey if you do
Best hosted swap on fees
Requesty — Model cost plus 5%, 600+ models, bring-your-own-keys, spend limits and EU data residency, with no server to maintain.
Best self-hosted swap for governance
Bifrost — Apache-2.0 and free to self-host, with virtual keys, budgets and failover across 23+ providers. Latency figures are the vendor's own.
Best when you also want guardrails and logs
Portkey — The MIT gateway is free to self-host; the hosted Production plan is $49 a month for 100k recorded logs.
None of these is an affiliate partner, so this pick is not commission-driven.
Quick answer
The LiteLLM alternatives split by who runs the gateway. Hosted, so you run no server: OpenRouter gives 500+ models behind one key with a 5.5% platform fee on the Standard plan; Requesty charges the model cost plus 5% and supports bring-your-own-keys. Self-hosted and free like LiteLLM: Portkey's MIT-licensed gateway adds guardrails and optional hosted logs ($49 a month for 100k recorded logs), Bifrost is Apache-2.0 with virtual keys, budgets and failover across 23+ providers, OmniRoute pools the free tiers of 90+ providers, and 9Router gives one local endpoint for coding agents. Cloudflare AI Gateway is hosted but free for analytics, caching and rate limiting in front of your own provider keys. LiteLLM itself stays the broadest free option, with 100+ providers, virtual keys, budgets and spend tracking.
Alternatives at a glance
Same job, different trade-offs. Open any profile for full sources and limitations.
| Tool | Category | Best for | Pricing note | Pricing source |
|---|---|---|---|---|
| OpenRouter | AI Gateway | Developers and teams who want one API key for many models with hosted reliability, without running or maintaining any gateway infrastructure. | Pay-as-you-go with prepaid credits; model prices pass through at list price. Free: 25+ free models from 4 providers, 50 requests/day. Standard: 500+ models from 80+ providers, 5.5% platform fee, provider passthrough rate limits. Business: 8% platform fee, adds EU/US in-region routing and up to 1,000 workspaces. BYOK is fee-free up to $25,000/month of list-price inference, then 5%. Enterprise: custom, SSO/SAML and SLAs. | openrouter.ai · checked |
| Requesty | AI Gateway | Developers who want an OpenRouter-style hosted gateway with a flat 5% markup, bring-your-own-keys and EU data residency, without running infrastructure. | Free: $0, access to free models, 200 requests/day, routing, caching and EU data residency. Pay as you go: model cost plus a 5% markup (a $10 per 1M-token model costs $10.50), 600+ models, bring your own keys, spend limits, MCP gateway. Enterprise: custom (SSO, RBAC, audit logs, SLAs). No per-seat pricing or minimum spend. | requesty.ai · checked |
| Portkey | AI Gateway | Teams putting LLM features into production that need observability, guardrails and key management on top of routing, with the option to self-host the open-source gateway. | Developer: free forever, 10k recorded logs/month, 3-day log retention, not for production. Production: $49/month, 100k recorded logs/month (+$9 per additional 100k), 30-day log retention, LLM guardrails, RBAC, semantic caching. Enterprise: custom (SSO, VPC hosting, SOC2/HIPAA). The open-source gateway is free to self-host. | portkey.ai · checked |
| Bifrost | AI Gateway | Engineering teams running high-traffic LLM apps that want a fast, self-hosted gateway with failover and caching, and an enterprise path for clustering and guardrails. | Open source under Apache 2.0 and free to self-host. Bifrost Enterprise (private deployments with adaptive load balancing, clustering, guardrails and an MCP gateway) has a 14-day free trial; its price is not published. | getmaxim.ai · checked |
| Cloudflare AI Gateway | AI Gateway | Developers already on Cloudflare who want free caching, rate limiting and usage analytics in front of their existing provider keys with one line of code. | Core features (dashboard analytics, caching, rate limiting) are free on all plans. Unified Billing adds a 5% fee on purchased credits, with provider inference passed through at no markup. DLP scanning is free; Guardrails is billed as Workers AI inference. Logs: gateways created from September 24, 2026 follow Workers Logs pricing; earlier gateways keep Legacy Logs (100,000 logs on Workers Free, 10 million per gateway on Workers Paid). Logpush needs Workers Paid (10M requests/month, then $0.05 per million). | developers.cloudflare.com · checked |
| OmniRoute | AI Gateway | Developers and creators who want free or low-cost access to many LLM providers without juggling separate API keys, subscriptions, and rate limits. | Free and open source (MIT). Self-hosted, no subscription; free tiers of 90+ providers are reachable through one endpoint. Check the official site for current provider terms and limits. | github.com · checked |
| 9Router | AI Gateway | Developers who want one local endpoint for Claude Code, Codex, Cursor, Cline, Copilot, OpenCode, and other compatible tools, with provider fallback and a dashboard instead of manual endpoint switching. | Free and open source under the MIT license; self-hosted through the npm package or source. The gateway has no subscription price, but connected providers, accounts, quotas, and their Terms of Service still determine the real cost and availability. Verify live limits before relying on a provider. | 9router.com · checked |
Choose the right alternative
OpenRouter
Developers and teams who want one API key for many models with hosted reliability, without running or maintaining any gateway infrastructure.
Pay-as-you-go with prepaid credits; model prices pass through at list price. Free: 25+ free models from 4 providers, 50 requests/day. Standard: 500+ models from 80+ providers, 5.5% platform fee, provider passthrough rate limits. Business: 8% platform fee, adds EU/US in-region routing and up to 1,000 workspaces. BYOK is fee-free up to $25,000/month of list-price inference, then 5%. Enterprise: custom, SSO/SAML and SLAs.
Requesty
Developers who want an OpenRouter-style hosted gateway with a flat 5% markup, bring-your-own-keys and EU data residency, without running infrastructure.
Free: $0, access to free models, 200 requests/day, routing, caching and EU data residency. Pay as you go: model cost plus a 5% markup (a $10 per 1M-token model costs $10.50), 600+ models, bring your own keys, spend limits, MCP gateway. Enterprise: custom (SSO, RBAC, audit logs, SLAs). No per-seat pricing or minimum spend.
Portkey
Teams putting LLM features into production that need observability, guardrails and key management on top of routing, with the option to self-host the open-source gateway.
Developer: free forever, 10k recorded logs/month, 3-day log retention, not for production. Production: $49/month, 100k recorded logs/month (+$9 per additional 100k), 30-day log retention, LLM guardrails, RBAC, semantic caching. Enterprise: custom (SSO, VPC hosting, SOC2/HIPAA). The open-source gateway is free to self-host.
Bifrost
Engineering teams running high-traffic LLM apps that want a fast, self-hosted gateway with failover and caching, and an enterprise path for clustering and guardrails.
Open source under Apache 2.0 and free to self-host. Bifrost Enterprise (private deployments with adaptive load balancing, clustering, guardrails and an MCP gateway) has a 14-day free trial; its price is not published.
Cloudflare AI Gateway
Developers already on Cloudflare who want free caching, rate limiting and usage analytics in front of their existing provider keys with one line of code.
Core features (dashboard analytics, caching, rate limiting) are free on all plans. Unified Billing adds a 5% fee on purchased credits, with provider inference passed through at no markup. DLP scanning is free; Guardrails is billed as Workers AI inference. Logs: gateways created from September 24, 2026 follow Workers Logs pricing; earlier gateways keep Legacy Logs (100,000 logs on Workers Free, 10 million per gateway on Workers Paid). Logpush needs Workers Paid (10M requests/month, then $0.05 per million).
OmniRoute
Developers and creators who want free or low-cost access to many LLM providers without juggling separate API keys, subscriptions, and rate limits.
Free and open source (MIT). Self-hosted, no subscription; free tiers of 90+ providers are reachable through one endpoint. Check the official site for current provider terms and limits.
9Router
Developers who want one local endpoint for Claude Code, Codex, Cursor, Cline, Copilot, OpenCode, and other compatible tools, with provider fallback and a dashboard instead of manual endpoint switching.
Free and open source under the MIT license; self-hosted through the npm package or source. The gateway has no subscription price, but connected providers, accounts, quotas, and their Terms of Service still determine the real cost and availability. Verify live limits before relying on a provider.
Direct comparisons
Why teams look past LiteLLM
The software is free, but the operation is yours: the catalog records PostgreSQL plus optional Redis, upgrades and uptime as your responsibility, and governance features such as SSO/SCIM and audit logs need the paid Enterprise edition.
There is also a supply-chain lesson. On March 24, 2026, PyPI releases 1.82.7 and 1.82.8 of the litellm package were published with credential-stealing malware, live for about 40 minutes. If you stay on LiteLLM, pin versions and verify checksums; if you switch, apply the same discipline to whichever gateway you pick.
Hosted or self-hosted: what changes
Hosted gateways trade a fee for convenience. OpenRouter lists a 5.5% platform fee on Standard (8% on Business) with provider prices passed through; Requesty adds 5% to every paid request and has a free plan limited to free models at 200 requests a day. Neither needs a server from you, but requests pass through their infrastructure.
Self-hosted gateways charge no gateway fee, the same as LiteLLM, and you keep requests on your own machine. You pay providers directly and carry the operations. Cloudflare AI Gateway sits in between: hosted, free core features, and your provider keys stay yours.
Which alternative fits which job
Portkey: the closest feature match if you want production tooling. Its open-source gateway (MIT) is free; the hosted plans meter recorded logs, with Developer free at 10k logs a month (not for production) and Production at $49 a month for 100k logs, plus $9 per additional 100k.
Bifrost: a self-hosted Apache-2.0 gateway aimed at high-traffic apps, with failover, load balancing, semantic caching, virtual keys and OpenTelemetry. The benchmark of about 11 microseconds of overhead per request at 5,000 requests a second is Bifrost's own, so measure with your traffic. Clustering and guardrails are enterprise features.
Cloudflare AI Gateway: free analytics, caching and rate limiting on every Cloudflare plan. There is no self-hosted edition, and log pricing changed for gateways created on or after September 24, 2026.
OmniRoute and 9Router: free, self-hosted, and aimed at personal use and coding agents rather than team budgets. OmniRoute pools free provider tiers; 9Router gives Claude Code, Codex and Cursor one local endpoint.
Run it 24/7 on your own server
LiteLLM and Portkey run on your own machine, so it stops answering when your laptop sleeps. A small Linux VPS with Docker keeps it online for scripts, agents and teammates, with a fixed monthly cost instead of per-token markup.
- Start with the smallest plan that runs Docker; upgrade only if you add users or local models.
- Keep the gateway bound to localhost or behind a firewall, and store provider keys as environment variables.
- Monthly billing lets you test before committing to a long term.
Some links are affiliate links. We may earn a commission if you buy through them. Payment never changes our rankings or editorial verdicts.
Frequently asked questions
- What is the best alternative to LiteLLM?
- It depends on whether you want to run a server. Without one, Requesty (model cost plus 5%) or OpenRouter (5.5% platform fee) are the closest hosted swaps. Self-hosted, Portkey's MIT gateway and Bifrost (Apache 2.0) are free like LiteLLM, with guardrails and logs on Portkey and virtual keys and budgets on Bifrost.
- Is there a free alternative to LiteLLM?
- Yes. Portkey's open-source gateway, Bifrost, OmniRoute and 9Router are free to self-host, and Cloudflare AI Gateway's analytics, caching and rate limiting are free in front of your own provider keys. Model usage is still billed by each provider.
- Is LiteLLM safe to use after the March 2026 incident?
- On March 24, 2026, PyPI releases 1.82.7 and 1.82.8 of the litellm package were published with credential-stealing malware and stayed live for about 40 minutes. Pin your version, verify checksums and follow the advisory in BerriAI/litellm issue 24518 before installing.
- What is the best alternative to LiteLLM?
- The most direct alternative to LiteLLM is OpenRouter. It is best suited for developers and teams who want one API key for many models with hosted reliability, without running or maintaining any gateway infrastructure. You can also evaluate Requesty, Portkey, Bifrost, Cloudflare AI Gateway, OmniRoute, 9Router depending on your specific requirements. Use the decision table above to compare pricing and feature trade-offs.
Keep the decision honest
Pricing, credit systems, and commercial terms change often. Every profile is checked against the official site — Verify the current plan before subscribing. Estimate your gateway cost If an alternative here does not fit, the comparisons hub covers more pairs.