Gateways that no longer exist

5 of the services routinely recommended as OpenRouter alternatives have shut down or pivoted out of the category, and 4 more were acquired or renamed. Most listicles still recommend them. Each entry below states what is actually there now, the evidence, and the date it was checked.

Unify

Pivoted away

unify.ai

Was:

A router that scored endpoints on cost, quality and speed and picked one per request.

Now:

A research lab working on symbolic reasoning and continual learning. There is no routing product.

Evidence:

unify.ai is a single-page research site; the pricing page returns 404.

Use insteadOpenRouterRequestyChecked September 1, 2026

Kluster.ai

Shut down

kluster.ai

Was:

Serverless open-model inference with a discounted batch tier.

Now:

Offline. The domain no longer serves a working product.

Evidence:

kluster.ai and docs.kluster.ai both redirect to www.kluster.ai, which fails to connect; DNS points at a redirect service.

Use insteadDeepInfraTogether AIChecked September 1, 2026

Predibase

Acquired

predibase.com

Was:

Fine-tuning and serving for small open-weight models.

Now:

Absorbed into Rubrik and folded into its agent cloud product.

Evidence:

predibase.com 301-redirects to rubrik.com/products/rubrik-agent-cloud.

Use insteadFireworks AIBasetenChecked September 1, 2026

Martian

Pivoted away

withmartian.com

Was:

A model router that claimed to pick the best model per prompt.

Now:

Operating as Thesean AI, an interpretability research lab. Its developer documentation is still online, but there is no publicly priced routing product.

Evidence:

withmartian.com presents Thesean AI; docs.withmartian.com remains reachable, so this is a pivot rather than a shutdown.

Use insteadOpenRouterVercel AI GatewayChecked September 1, 2026

Hyperbolic

Pivoted away

hyperbolic.ai

Was:

Low-cost serverless inference over open-weight models.

Now:

A GPU cloud. The token-billed inference pricing page is gone.

Evidence:

The pricing path returns 404 on both hyperbolic.ai and www.hyperbolic.ai.

Use insteadDeepInfraNovita AIChecked September 1, 2026

Anyscale

Pivoted away

anyscale.com

Was:

Serverless LLM endpoints priced per token.

Now:

Focused on Ray compute. The token-billed endpoint product is retired.

Evidence:

Anyscale's product surface now sells Ray compute rather than per-token endpoints.

Use insteadTogether AIFireworks AIChecked September 1, 2026

Lambda Inference API

Winding down

lambda.ai

Was:

A serverless Inference API over popular open models, alongside GPU rental.

Now:

Being wound down. Lambda continues as a GPU cloud.

Evidence:

Lambda states the Inference API is winding down on its own inference page.

Use insteadDeepInfraLambda GPU cloudChecked September 1, 2026

Nebius AI Studio

Renamed

nebius.com

Was:

Nebius AI Studio - EU-hosted serverless inference.

Now:

The same product, renamed Nebius Token Factory, on a new API host.

Evidence:

The API now lives at api.tokenfactory.nebius.com. Nothing was discontinued.

Use insteadNebiusChecked September 1, 2026

Helicone

Acquired

helicone.ai

Was:

An independent observability proxy for LLM calls.

Now:

Part of Mintlify. The proxy still operates.

Evidence:

Helicone announced joining Mintlify in 2026.

Use insteadPortkeyHeliconeChecked September 1, 2026

For the services that do still exist, see all 117 gateways.