Show HN: Distill and serve small models with frontier quality for half the cost

https://github.com/experientiallabs/world-model-optimizer

39•SilenN•54m ago

Hi HN, we built world-model-optimizer, an open-source tool for continually improve models specialized to agents. Today we are launching `wmo serve`, a tool to route repetitive tasks to distilled smaller models.

Agent traces you already capture are opportunities to get signal on how to make your model cheaper, faster, better. We continuously improve - your specialized model through distillation from open source models - model routing to frontier + custom models - token compaction to remove noise and save tokens

Demo: https://www.youtube.com/watch?v=2_m4Ze6mdko

Pass in traces and an OpenRouter key, and wmo starts a local OpenAI-compatible endpoint to run with your model at a lower cost with equivalent quality. Behind the scenes a router decides which tasks should go to the frontier versus your model. Tinker continually trains as new traces arrive.

We also offer a hosted solution for anyone that just wants a frontier quality endpoint with self-improvement over time at a 40%+ lower cost.

Comments

rglover•26m ago

Excited to play with this.

jack_pp•24m ago

Not sure I get it. The model you're improving is local? If so how do you even calculate cost compared to an API

SilenN•8m ago

Open source models.

wmo routes requests between frontier models and open source models that continuously train using Tinker. As the smaller models improve, more traffic gets routed to them.

Calculating cost is just tokens in/out.

yiyingzhang•22m ago

Cool idea! How do you guarantee privacy?

SilenN•7m ago

It's open source!

We do have a platform we'll be launching as well to manage training + serving for you which will require more diligent privacy guarantees.

digitaltrees•22m ago

Cool project

adrianco•21m ago

Local models need to be tuned to work well so this looks useful. Seems to be for general purpose model serving. I’ve been using https://github.com/adrianco/retort to run experiments for coding models across 13 different programming languages to see which frontier and local models work.

SilenN•15m ago

That's cool, thanks for sharing!

Art9681•14m ago

The absolute best way to prove this works is by releasing a model that was fine-tuned with this method and then showing benchmarks depicting the improvement delta between the base model and the fine tuned one.

The work is not done. Then release it to the masses and wait a few days for the actual real world anecdotes.

Until then, this is noise.

Reubend•10m ago

Yeah, this is just slop. No benchmarks, no concrete case studies, just some vibecoded "platform" to finetune models on your own traces.

Which is an idea that has some value, but also some weaknesses. And this implementation of it isn't forthcoming with that concept. You have to really dig in to understand what they're even talking about.

SilenN•4m ago

Happy to answer any qs.

irishcoffee•9m ago

Benchmarks are the ultimate consolidation of halnons razor.

teravor•8m ago

[flagged]

dang•7m ago

"Please don't post shallow dismissals, especially of other people's work. A good critical comment teaches us something."

https://news.ycombinator.com/newsguidelines.html

surround•3m ago

The title is misleading. This is model routing, not distillation.

I Championed Prediction Markets. Look What They've Become

Triton: DirectX 11 Driver for QEMU

Show HN: Simple HIIT timer for efficient workouts

How the Atmos Clock Functions

PGSimCity - How PostgreSQL Works

Show HN: Turn an image into a 3D blockbench model

DLC Protect – Open-Source ECDH Encryption for Steam DLCs

Liveness proofs in Veil: Part 1

Open coordination stack for lattice-building robot swarms

Quebec scraps AI and automation projects in the public sector

Odella [video]

Show HN: Blind-box commerce made easy

Unlocking the Halo: Campaign Evolved Secret Debug Menu

UBI-ticon: High Effort Feature Film using programmatic sonification

Open-source 3D-printed portable MRI machine built for under $70k

Show HN: Port Zero – how I learned to stop worrying and love PORT=0

Data centres in NZ: Jobs creators or destroyers?

Qwen 27B with local well writen tools just as powerful as cluade models?

From A.I. To the Deep State, Michel Foucault Foresaw It All

A Personal Note from Our Founder (2025)

Visualizing the Artemis II Mission

Ask HN: Is Zig worth it over Rust in 2026?

Most Googlebots Are Fake

Jeffy Loop, an autonomous code improvement loop for Claude Code

Fonts In Use – Find out where a font is used

From Soft to Hard Skills

Show HN: Lowkey Studio – Shader Powered visual effects compositor in the Browser

AI Moves into Family Life

How to self-host servers in your living room on static IPs

Show HN: Distill and serve small models with frontier quality for half the cost

Show HN: Distill and serve small models with frontier quality for half the cost

Comments

I Championed Prediction Markets. Look What They've Become

Triton: DirectX 11 Driver for QEMU

Show HN: Simple HIIT timer for efficient workouts

How the Atmos Clock Functions

PGSimCity - How PostgreSQL Works

Show HN: Turn an image into a 3D blockbench model

DLC Protect – Open-Source ECDH Encryption for Steam DLCs

Liveness proofs in Veil: Part 1

Open coordination stack for lattice-building robot swarms

Quebec scraps AI and automation projects in the public sector

Odella [video]

Show HN: Blind-box commerce made easy

Unlocking the Halo: Campaign Evolved Secret Debug Menu

UBI-ticon: High Effort Feature Film using programmatic sonification

Open-source 3D-printed portable MRI machine built for under $70k

Show HN: Port Zero – how I learned to stop worrying and love PORT=0

Data centres in NZ: Jobs creators or destroyers?

Qwen 27B with local well writen tools just as powerful as cluade models?

From A.I. To the Deep State, Michel Foucault Foresaw It All

A Personal Note from Our Founder (2025)

Visualizing the Artemis II Mission

Ask HN: Is Zig worth it over Rust in 2026?

Most Googlebots Are Fake

Jeffy Loop, an autonomous code improvement loop for Claude Code

Fonts In Use – Find out where a font is used

From Soft to Hard Skills

Show HN: Lowkey Studio – Shader Powered visual effects compositor in the Browser

AI Moves into Family Life

How to self-host servers in your living room on static IPs

Show HN: Distill and serve small models with frontier quality for half the cost