p.enthalabs

Show HN: We built open OpenRouter that turns usage into a better model

198 points · 17 visible top comments · 2026-08-27 21:18:35 UTC

github.com · Read Story HN original

Hi HN, we built an open source model gateway. It's a single place to manage our own self hosted, frontier, and open source models in one place.

It’s is rust native, built for concurrency, and implements all the config quirks across models and providers (streaming formats, tool calls, model parameters, rate limits, and different error behavior).

The gateway adds under 1 ms for BYOK requests and under 2 ms when Experiential supplies the provider key. It has every major inference provider, and 1000+ models refreshed daily via a codex agent that opens a PR.

Compared to other similar projects we’re open source, take no markup, allow you to mix local models with a marketplace, and use your traffic to (opt in) train you a model. Simple routing doesn’t warrant a 10% token markup.

The way we do this is given standardized OTel traces, we mine representative real tasks, use text world models to simulate rollouts for various models, apply an LLM judge, and fit a nearest neighbor classifier on top of an embedding of a prompt to decide the optimal model for each request. Usually this can map out a better pareto curve on cost/quality than just calling single models but it’s not perfect.

Using these simulations we can also do things like suggesting cache hit optimizations, new model suggestions, and training models.

It’s open source, so you can deploy it on your own infrastructure, use our hosted version with 0 markup, or read how we design for maximum availability on our website.

Comments

Areibman · 2026-08-27 22:12:13 UTC
Could you say more about how caching works? One major advantage of sticking with a single model is saving money on cached input tokens. I'd imagine if you swap between a bunch of models, you may improve performance but cost would would balloon out of control
purplecats · 2026-08-27 22:21:06 UTC
and caching is related to performance too ofc
SilenN · 2026-08-27 22:24:01 UTC
The trick is to rarely switch, or switch at task boundaries. Often the conclusion of routing is actually "this one model is actually at the pareto front for this task, just use it always".
cameronh90 · 2026-08-28 00:42:25 UTC
But then it's better to just not have a gateway switch models at all.

Just have the harness able to choose which model its sub-agents use, then tell it how to split up tasks and which models to use when doing so.

SilenN · 2026-08-28 00:46:29 UTC
That is another way to do. Or we can automatically figure out which models the subagents should be using for you. And update them as new models come out and the work your subagents do changes. More than one way to skin a cat.
aerzen · 2026-08-28 10:38:20 UTC
This does make sense. I generally only switch between models in pi when creating a new session. And it is apparent from the promt if this just a "how to see open ports on linux" or "make a concrete plan for feature X"
ashermania · 2026-08-27 22:14:48 UTC
Finally an open source tool doing this!
23david · 2026-08-27 22:47:54 UTC
Super interesting and congrats on the release. Curious if you initially had this in Python and then rewrote in Rust?
SilenN · 2026-08-27 22:50:53 UTC
Yep! If you look at the commit history that's exactly what happened.
cheema33 · 2026-08-27 23:54:09 UTC
I have not tried it yet. Is it similar to LiteLLM? If so, what sets it apart?
kfallah15 · 2026-08-28 00:11:56 UTC
Router and model optimization from traffic is the main differentiator
SilenN · 2026-08-28 00:48:13 UTC
Also a hosted marketplace, not just BYOK
ceroxylon · 2026-08-28 00:58:42 UTC
>The gateway adds under 1 ms for BYOK requests

Amazing! Really brilliant idea, thank you for sharing this project. There is so much ground to cover in the LLM gateway / routing / reporting world, and this is a great start. The Tinker implementation is my favorite part, fine tuning is much better than a sea of context files.

kfallah15 · 2026-08-28 01:21:19 UTC
Thanks! We are going to add continual RL via Tinker soon too
akshay_akula · 2026-08-28 01:03:44 UTC
Open source and no markup is the right default for a gateway. The caching question above is the one I would want answered before swapping models though.
SilenN · 2026-08-28 01:27:31 UTC
Ans: we rarely switch, often times it's just a "switch to using this model for your agent"
akshay_akula · 2026-08-28 13:19:11 UTC
You guys should look into ngrok ai gateway. We have some small models running on local hardware that we tried to use but it was too painful. Just wanted to not waste all the compute hit one endpoint and be done but as a team.
0xbadcafebee · 2026-08-28 01:42:26 UTC
You started it a week ago? I look forward to checking back in 3 weeks when you've exited for $1B
SilenN · 2026-08-28 02:06:57 UTC
See you soon
rdslw · 2026-08-28 07:12:11 UTC
impressive only if using pre-gpt era assumptions about saas/products/software.

unfortunately a small team can reproduce it in two months, which greatly lowers value of it.

we, as a collective, have to change our value-judging logic and tune it to post AI world.

SilenN · 2026-08-28 07:36:50 UTC
Thanks for the positivity tyre! If you look at our git history, we pivoted and only started building the gateway recently. Before that we were building research infrastructure that now powers the intelligence features we provide.
swthbht · 2026-08-28 03:03:27 UTC
Very cool. Does your gateway decide effort levels as well? Or just models?
SilenN · 2026-08-28 03:22:07 UTC
Yep! One interesting example is often Opus 5 on low reasoning ~= Opus 5 on high reasoning.
sangwook · 2026-08-28 03:48:22 UTC
What online signal recalibrates simulated rankings against actual task success? Also do you have a plan to support semantic caching at the router level?
kfallah15 · 2026-08-28 04:05:25 UTC
For the online signal, we use a LLM judge with a rubric calibrated offline by the user via TUI. UX of the calibration is a major focus area. Semantic caching is interesting, open to supporting it but not currently planned.
forgetme2020 · 2026-08-28 03:49:06 UTC
what's the business model here. How does experiential labs make money
kakugawa · 2026-08-28 03:58:11 UTC
They make money on enterprise plans: https://www.experientiallabs.ai/pricing#enterprise

Look at the Intelligence features in the Enterprise plan:

* Per-prompt model optimization

* Caching

* A model you own, trained on your traffic

kfallah15 · 2026-08-28 04:08:06 UTC
yep, it will be through enterprise licenses and our own hosted platform built on the repo
rdslw · 2026-08-28 07:17:48 UTC
the business model, I suspect, is classic rug-pull in some time after building user base.

proof: boldly claiming being open source in literally first sentence, while cowardly hiding on-by-default telemetry (WTF??) in truly last paragraph of readme.

sorry to sound harsh, but this is typical old era playbook here.

in the era of AI, fortunately, such products has much lower value. people and VCs didn’t yet tune to it.

SilenN · 2026-08-28 07:34:00 UTC
Telemetry is off by default. PostHog is for usage analytics on the open source repo. Audit it if you're skeptical.

We make money off enterprise licenses and hosting models.

I have strong reason to suspect you either can't read or are a bad actor.

aHumbleUser · 2026-08-28 13:51:54 UTC
Your GitHub readme says telemetry is on by default
gpiechnik2 · 2026-08-28 08:05:38 UTC
great design! i love it
SilenN · 2026-08-28 08:39:51 UTC
Thank you!
nejch · 2026-08-28 08:46:43 UTC
A part of me wishes the open source community would focus making research and industry-backed initiatives like the vLLM Semantic Router rock solid. Then I'd spend less time every month checking if this or that new model router has differentiating over vllm-sr :)

At least for open source inference, it seems like there's healthy competition centered around vllm/sglang, but 2026 seems to be for model routers what 2025 was for agent harnesses.

d2p · 2026-08-28 09:09:34 UTC
> and use your traffic to (opt in) train you a model.

Is there more info on this? I'm curious exactly what it is. Is it fine-tuning/LoRA on some base model? Don't cloud providers encrypt reasoning now - does that prevent this?

croemer · 2026-08-28 09:27:03 UTC
Probably shouldn't call it "Open router" in the title as that's a specific brand. Maybe you meant "we built something like OpenRouter".

Rereading I see you wrote "open OpenRouter", which looks a bit like a typo at first glance.

foremerge · 2026-08-28 13:14:25 UTC
We built our own model router at GPTree but this looks interesting. Caching is definitely one of the hardest parts to get right, especially in our scenario where users can branch from any part of the conversation and keep the dynamic context from the parent (unlike most other platforms that lock in the context once you branch).
jakswa · 2026-08-28 13:20:22 UTC
I think I have to look at setting up one of these AI gateways for work, since AWS bedrock is such a PITA to hook a harness up to over IAM roles. Also I still can't believe bedrock hasn't released any open models in months (so there's paranoia that I'll want to swap in another provider).

Really though I'm hoping anthropic fixes the oppressive claude verbosity. I saw someone refer to being "clauderboarded" and my brain cannot let go of this as Claude's tokens bombard me.