LiteLLM is the one to build on: it ships every few days and puts budgets in the free build, while Portkey's open-source gateway has not had a release since January and now belongs to Palo Alto Networks.
Choose LiteLLM if several teams share provider keys and you need per-team budgets, spend reports and fallbacks you run yourself, from a project that is actively maintained.
Choose Portkey's gateway if you need a light router on Cloudflare Workers or another edge runtime, or you already buy Palo Alto's Prisma AIRS and want the hosted version.
Free and open source under MIT, except the enterprise/ folder, which is under a separate commercial licence. Enterprise is custom-priced with a free 30-day trial key.
Free and open source under MIT. Hosted plans: Developer free (10,000 recorded logs a month, 3-day log retention); Production $49/month (100,000 logs, then $9 per 100,000); Enterprise custom.
Best for
Platform teams that must give many developers governed access to many model providers from their own infrastructure.
Teams on edge runtimes, or enterprises already buying Palo Alto security, who want a light routing layer.
Language and runtime
Python proxy and SDK; Docker, Helm, Terraform
TypeScript on Node, Bun, Docker or Cloudflare Workers
Coverage claim
100+ providers
1,600+ models, per the README
Budgets and spend tracking
Virtual keys, budgets and rate limits in the free build
In hosted plans, not the open-source gateway
Licence
MIT, except an enterprise/ folder under a commercial licence
MIT, no enterprise folder
Latest release
v1.106.4 on 25 October 2026
v1.15.2 on 12 January 2026
Paid option
Enterprise, priced on request, with a 30-day trial key
Hosted Production $49 a month for 100,000 logs; Enterprise custom
Owner
BerriAI
Palo Alto Networks, since 29 May 2026
Security record
Poisoned PyPI releases in March 2026; 17 advisories this year
One medium-severity SSRF advisory, December 2025
Both are MIT-licensed gateways that put many providers behind one OpenAI-style API with fallbacks and retries. LiteLLM is a Python proxy with virtual keys, budgets and spend tracking in the open-source build. Portkey is a small TypeScript router that runs on edge runtimes such as Cloudflare Workers, but keeps logs, budgets and analytics in its hosted plans, and its repository has been quiet since May 2026.
These were once the two obvious open-source gateways, and the gap between them is now about maintenance more than features. LiteLLM released v1.106.4 on 25 October 2026 and has about 62,000 GitHub stars. Portkey's gateway, with about 13,300, last released v1.15.2 on 12 January; its default branch has not changed since 25 May, four days before Palo Alto Networks completed its acquisition of Portkey. The Gateway 2.0 branch meant to fold the enterprise gateway into open source has been idle since March.
The products also draw the open-source line in different places. LiteLLM's free build issues virtual keys, tracks spend by key, user and team, and enforces budgets, which is the main reason a platform team runs a gateway at all. Portkey's open-source build routes, retries and falls back from one JSON config, while logs, semantic caching, analytics and most of its pre-built guardrails sit in the hosted product. That hosted product is cheap to start, $49 a month for 100,000 logs plus $9 per extra 100,000, but it means your request logs live with a vendor now sold as Prisma AIRS.
Portkey keeps one real technical advantage. Its small TypeScript core runs on Cloudflare Workers, where a Python proxy cannot, and the README claims under 1 ms of added latency against LiteLLM's 8 ms at P95; both figures are the projects' own. Its security record is also lighter, a single medium-severity SSRF advisory against LiteLLM's 17 advisories in 2026 and the March supply-chain attack on its PyPI releases. Some of that difference reflects how much code each project has shipped this year.
For a team choosing today, LiteLLM is the one that will support next month's model; a Portkey gateway with no release in nine months may need passthrough for providers added since. The comparison would change if Palo Alto ships Gateway 2.0 as open source, and that is the signal worth watching on Portkey's release page.
Which should you choose?
The jobs each one is best at, from our reviews.
Pick LiteLLM if you are…
A platform team giving 300 engineers access to OpenAI, Anthropic and Bedrock
Issue per-team keys with budgets and report spend by team
Run the proxy on Kubernetes with Postgres and Redis, hand each team a virtual key with a budget, and give finance spend by key, user and team. Free in the open-source build; SSO for the admin UI needs an enterprise licence.
An SRE responsible for an LLM-backed product's uptime
Fail over from Azure OpenAI to OpenAI when one deployment degrades
Put both deployments in one model group with fallbacks and retries in the router. The cost is an extra network hop and a component that holds every provider key, so pin a -stable image and verify its cosign signature.
A head of engineering at a regulated fintech
Decide between self-hosted open-source LiteLLM and the enterprise licence
Enterprise is custom-priced with a 30-day self-hosted trial key and adds SSO, SCIM, audit logs and a one-hour Sev 0 response target. Either way, budget time for patching: four critical advisories landed in 2026.
Pick Portkey AI Gateway if you are…
A developer running an app on Cloudflare Workers
Put OpenAI and Anthropic behind one endpoint with automatic fallback
Deploy the gateway to Workers and describe the fallback in a config. It costs nothing beyond Workers usage; with no release since January, very new models may need passthrough rather than a native provider.
A five-engineer startup with a production LLM feature
Get request logs and cost dashboards without running an observability stack
Hosted Production at $49 a month covers 100,000 logged requests with 30-day log retention; at one million requests a month the bill is about $130, $49 plus nine extra blocks at $9.
A security team at an enterprise already on Prisma AIRS
Govern which models and agents staff can reach, centrally
Enterprise or Prisma AIRS terms are custom-priced; Enterprise lists SSO, SOC 2 Type 2, GDPR, HIPAA with custom BAAs, and VPC deployment.
LiteLLM vs Portkey AI Gateway pricing
LiteLLM
Free and open source under MIT, except the enterprise/ folder, which is under a separate commercial licence. Enterprise is custom-priced with a free 30-day trial key.
Portkey AI Gateway
Portkey AI Gateway plans
Open source
FreeMIT
Developer
Free10,000 recorded logs a month, 3-day log retention
Production
$49/month100,000 logs, then $9 per 100,000
Enterprise
Custom
Pros and cons
LiteLLM
Good
100+ providers in OpenAI format
Virtual keys, budgets and fallbacks in the free build
Cosign-signed -stable images with 12-hour load tests
Terraform modules for AWS and GCP
Not so good
17 security advisories in 2026, four critical
Poisoned PyPI releases in March 2026
Open-core boundary is unclear in the docs
Enterprise pricing not published
Portkey AI Gateway
Good
MIT licence with no enterprise folder
Small TypeScript core that runs on Cloudflare Workers
One config for fallbacks, retries and conditional routing
Hosted logs from $49 a month
Not so good
No release since January 2026; 2.0 branch idle since March
Roadmap now set by Palo Alto Networks
Caching, analytics and most guardrails are hosted-only
SSO and compliance certifications only on Enterprise
Before you commit
LiteLLM vs Portkey AI Gateway: setup, integrations, data, cost and limits
Topic
LiteLLM
Portkey AI Gateway
Setup
uv tool install 'litellm[proxy]' then litellm --model gpt-4o runs a local proxy on port 4000; production uses the Docker image, the Helm chart or Terraform modules for AWS (ECS Fargate, Aurora, ElastiCache) and GCP (Cloud Run, Cloud SQL).
`npx @portkey-ai/gateway` runs it locally; the deployment guide covers Docker, Bun, Node and Cloudflare Workers.
Integrations
OpenAI-compatible /chat/completions, /responses, /embeddings, /messages, /batches and more across 100+ providers; logging to Langfuse, MLflow, Helicone and others; MCP and A2A gateways; a Python SDK that works without the proxy.
OpenAI-style API in front of 1,600+ models, per the README; configs for routing and guardrail hooks; the hosted product adds an MCP gateway, prompt management and analytics.
Data and privacy
Self-hosted, so prompts go only to the providers you configure; the production guide mentions no telemetry to BerriAI. The proxy keeps keys, spend logs and optionally prompts in your Postgres. Seventeen security advisories were published in 2026, four critical.
Self-hosted, requests pass only through your instance. Hosted logs are kept 3 days on Developer, 30 days on Production and for custom periods on Enterprise; SOC 2 Type 2, GDPR and HIPAA are listed for Enterprise. One medium-severity SSRF advisory (December 2025).
Cost at scale
The docs size each worker at 1 vCPU and 4 GiB, so an 8-worker node needs 8 vCPU and 32 GiB, plus Postgres and Redis 7 once you run more than one instance. Enterprise is priced on request.
Open source is free. Hosted Production is $49 a month for 100,000 logs plus $9 per additional 100,000, so a million requests is about $130 a month; the comparison table caps Production at 3 million, beyond which Enterprise pricing is custom.
Limits
5,748 open issues and pull requests; releases every few days mean frequent upgrades; Postgres connection pools must be sized per instance and worker; the Rust gateway is still in beta.
No release since 12 January 2026 and no commit since 25 May; 307 open issues; semantic caching and analytics are not in the open-source build.
Common questions
Is LiteLLM better than Portkey AI Gateway?
LiteLLM is the one to build on: it ships every few days and puts budgets in the free build, while Portkey's open-source gateway has not had a release since January and now belongs to Palo Alto Networks. We rate LiteLLM 4.3 and Portkey AI Gateway 3.7 out of 5.
When should I choose LiteLLM?
If several teams share provider keys and you need per-team budgets, spend reports and fallbacks you run yourself, from a project that is actively maintained.
When should I choose Portkey AI Gateway?
If you need a light router on Cloudflare Workers or another edge runtime, or you already buy Palo Alto's Prisma AIRS and want the hosted version.
How much do LiteLLM and Portkey AI Gateway cost?
LiteLLM: Free and open source under MIT, except the enterprise/ folder, which is under a separate commercial licence. Enterprise is custom-priced with a free 30-day trial key. Portkey AI Gateway: Free and open source under MIT. Hosted plans: Developer free (10,000 recorded logs a month, 3-day log retention); Production $49/month (100,000 logs, then $9 per 100,000); Enterprise custom.
