Portkey vs Langfuse in 2026 - Gateway or Observability Platform?

People search “Portkey vs Langfuse” expecting a straight fight, but they are not the same kind of tool. Portkey is an LLM gateway - it routes your calls to 1,600+ models with fallbacks, caching, budgets and guardrails, and observability rides along as a bundled feature. Langfuse is a dedicated observability platform - tracing and evals are the whole point. So the real question is not “which is better,” it is “which problem do you have.” And for a lot of teams, the answer is both, running together.

I have added Helicone as a reference point, because it was the other proxy-based tool people compared here - though as you will see, it is no longer a live option.

The short version

Tool	Best for	Starting price	Self-host	Core job
Portkey	Multi-provider routing, budgets, guardrails	Free / $49/mo	Gateway only, not observability	Gateway
Langfuse	Deep tracing and evals, cheap at scale	Free / $29/mo	Free, MIT, near-complete	Observability
Helicone	Nobody new - maintenance mode	Free / $79/mo	Free, Apache-2.0, but frozen	Proxy (frozen)

Portkey: a gateway first, observability second

Portkey sits between your app and your model providers as a drop-in proxy. One base URL change routes your existing OpenAI SDK calls through it, and on top of routing it adds fallbacks, load balancing, retries, timeouts, caching, budgets, rate limits and 50+ guardrails. If your problem is “we call five providers, costs are exploding, and we need fallbacks and budgets in one place,” Portkey is aimed squarely at you. Its OpenTelemetry support is strong - it ingests OTel and enriches traces with cost and token metrics per the GenAI conventions. It is well-backed, with a $15M Series A led by Elevation Capital, and the open-source gateway repo sits at roughly 12,500 stars.

The gotcha is where the value migrates. The open-source gateway is Apache 2.0 and self-hosts free - but that gets you routing and a basic dashboard, not real observability. Logs, traces, analytics and retention live on the managed Production tier at $49/mo for 100k logs. And the meter caps logs, not requests - past the cap your traffic keeps flowing but your observability quietly goes dark. So a team that self-hosts Portkey expecting a full observability platform gets a proxy and a basic dashboard, and discovers the analytics they wanted are a paid feature. One more caveat - Portkey markets sub-1ms added latency, but a Kong-run benchmark reported Portkey about 65% higher latency than Kong’s gateway. Both are interested parties, so test on your own path.

Langfuse: the observability platform Portkey is not

Langfuse is the open-source default for LLM observability. It does tracing, evals, prompt management and human annotation, and unlike Portkey it gives you the whole product when you self-host. In the self-hosted build, only three features are enterprise-gated - tracing, evals, prompt management and RBAC are all free under MIT. It is framework-agnostic and runs as an OpenTelemetry backend on an OTLP endpoint. The economics are the headline - at 1M events a month it runs about $101/mo managed, roughly 25x cheaper than LangSmith for comparable volume, and self-hosting removes the per-trace cost entirely.

Its catch is operational, not commercial. Langfuse v3 needs Postgres plus ClickHouse, Redis and S3-compatible storage - four services - and the migration to that architecture is where self-hosters get stuck. gRPC OTLP is not supported yet, HTTP only. And it is now a ClickHouse subsidiary after the January 2026 acquisition, worth filing away for a multi-year bet. What Langfuse does not do is route your traffic - it is the observability backend, not the gateway.

Helicone: the frozen third option

Helicone was the other proxy-based tool people weighed here, and it did lightweight logging well - a zero-code base-URL swap in any language. But Mintlify acquired it on 3 March 2026 and put it in maintenance mode - security and bug fixes only, no roadmap, and Mintlify is helping customers migrate off. For a new buyer that ends the conversation. If you liked Helicone’s proxy convenience, Portkey is the live gateway to look at; for observability depth, Langfuse.

Portkey vs Langfuse: which should you pick?

• You call multiple LLM providers and need routing, fallbacks, budgets and guardrails in one place - Portkey, and expect to pay for the managed tier once you need real logs.
• You need deep tracing and evals and want to self-host the whole thing free - Langfuse, if you can run the four-service stack.
• You want the cheapest observability at production trace volume - Langfuse. Portkey’s observability is a paid add-on to a gateway.
• You have both problems - run both. Portkey as the gateway, Langfuse as the observability backend fed over OpenTelemetry. That pairing is the right mental model, and Portkey’s OTel enrichment makes it clean.
• You want a live proxy-based tool - Portkey. Helicone is frozen and off the table for new builds.

The honest framing: this is not a case where one wins. If you genuinely only need one, pick by your actual pain - spend and routing chaos points to Portkey, thin visibility into model behavior points to Langfuse. Most teams at scale end up with a gateway and a dedicated observability backend anyway.

For the wider field, see our Portkey alternatives and Langfuse alternatives breakdowns, or the head-to-head Portkey vs Helicone. Every price and date here was read from each vendor’s own materials on 26 July 2026. This category ships breaking changes monthly, so we re-verify every 30 days.