How much does LiteLLM cost?
LiteLLM's open-source gateway is free to self-host, and LiteLLM Enterprise is an annual license sized to your gateway request capacity, deployment and support needs, "never per token", with no public price. Either way you pay every model provider separately and run the gateway yourself. Standard Compute is the hosted, cost-optimized alternative for coding agents: a flat plan from $19 with compute included and routing that picks an efficient model per request.
LiteLLM free vs Enterprise
| Open source | Enterprise | |
|---|---|---|
| Price | $0, "Free forever" | Annual license sized to request capacity, deployment and support; talk to sales |
| Hosting | Self-hosted | Self-hosted, air-gap available |
| Virtual keys, budgets, rate limits | Included (needs a database) | Included, plus budgets per model on a key, tag budgets and soft-budget alerts |
| SSO | Free for up to 5 users | Any number of users, plus SCIM and JWT auth |
| Audit logs, RBAC, IP allowlists, key rotation | Not included | Included |
| Support | GitHub and community | 24/7 support and SLAs |
| Trial | Not needed | Instant 30-day trial key by email |
From litellm.ai/pricing and docs.litellm.ai/docs/enterprise, checked October 13, 2026. Enterprise can be bought direct or through AWS Marketplace.
What a coding-agent month costs on LiteLLM
Pinned to one Claude model, a typical month costs less on LiteLLM and a busy month less on a Standard Compute plan, which covers about 19 of its 20 days. LiteLLM adds nothing to the token price, so its bill is your providers' bill. What it changes is control: virtual keys, budgets and routing in your own infrastructure. A hundred typical coding-agent requests cost $5.33 on Claude Sonnet 5.5 and $0.41 on GPT-6 Luna, so which model handles each step still decides the month.
| Month of coding-agent use | Provider bill through LiteLLM | Standard Compute plan |
|---|---|---|
| Typical days, Claude Sonnet 5.5 | $71 + hosting | Standard, $89 with a $95 budget |
| Typical days, Claude Opus 5.5 | $141 + hosting | Pro, $249 with a $269 budget |
| Busy days, Claude Sonnet 5.5 | $579 + hosting | Pro Plus, $499 with a $549 budget (about 19 busy days) |
| Typical days, GPT-6 Luna | $5 + hosting | Starter, $19 with a $20 budget |
Month = 20 working days at Anthropic and OpenAI list prices. Typical is the median active day in our Agent Burn Index, busy the 90th percentile, from coding agents such as OpenCode, Kilo and Roo on our gateway, September 13 to October 12, 2026; Claude Code and Codex were not in the sample. Hosting means the server and Postgres database you run, plus the time to operate them.
Costs that don't show on the pricing page
- Budgets need a database. LiteLLM's docs: "none of them cap anything on a DB-less deployment", and the global max_budget fails open there.
- Routing is yours to configure. Cost-based routing picks the cheapest deployment of one model, and the beta complexity router sends requests to SIMPLE, MEDIUM, COMPLEX and REASONING tiers you fill in.
- Upgrades and security are yours. In March 2026 two PyPI releases (1.82.7 and 1.82.8) shipped a credential stealer; LiteLLM deleted them and said Docker image users were not affected. Pin versions and read release notes.
- Each provider needs its own account, key, billing and rate-limit tier.
The catch
Free software is not a free gateway. You pay providers at list price, run a server and database, and own the uptime, upgrades and security of the thing every agent request goes through.
Where Standard Compute fits
Standard Compute is the hosted, cost-optimized option for coding agents: a flat plan from $19 a month with included compute that stops at its budget, smart routing that sends routine steps to efficient models, a plan per key for teams and agencies, Claude Code on /v1/messages and Codex on /v1/responses, and Europe or US rules on any plan. It also works behind LiteLLM as an OpenAI-compatible provider.
FAQ
Is LiteLLM free?
The open-source gateway is free to self-host, with virtual keys, budgets and routing. You still pay each model provider for tokens and pay to host the gateway.
How much is LiteLLM Enterprise?
LiteLLM doesn't publish a price. Enterprise is an annual license sized to your gateway request capacity, deployment architecture and support needs, never per token, and a 30-day trial key is emailed on request.
What's the difference between LiteLLM free and Enterprise?
Enterprise adds SSO beyond five users, SCIM, audit logs, RBAC, IP allowlists, key rotation, secret managers, budgets per model on a key, and 24/7 support with SLAs. The core gateway is the same image with a license key.
Is there a hosted LiteLLM cloud?
Not on LiteLLM's site today: both plans are self-hosted. If you want a managed endpoint with budgets per key, a hosted API such as Standard Compute fills that gap.
Does LiteLLM add a markup on tokens?
No. You call providers with your own keys and they bill you directly at their prices.