Today I made a Northwind database demo with Hy3:

http://1y1z.com:4832

It has all the typical bells and whistles:

• full CRUD editing of all fields in the database (create, read, update, delete)
• complete sort and filter functionality in all datatable displays (click column headers to sort)
• auth login to save changes to the database ('admin'/'admin')
• input/export CSV files to/from every table
• export order invoices to PDF
• an analytics page with charts
• responsive layout for mobile and desktop screens
• toggle UI between dark/light mode

Hy3 produced great output for every software development prompt I gave it, and for all the agentic tasks I tried with it. It built everything quickly without any hitches, and didn't choke for a moment. It's also done amazingly well with the suite of world knowledge tests I run on every new model. For a model in the several hundred billion parameter class, which can run on a single machine with 128GB RAM (DGX Spark, Strix Halo, and PCs with 4x 32GB GPUs), I haven't seen anything better yet.

So, this model is certainly going to be among my top 3 for local hosting.

My next tests will be with clustered DGX Sparks. From what I understand, vLLM already supports Hy3, with fast tensor parallelism. That means, using two Asus GX10s connected by the ConnectX ports, the expectation is for speeds to be about 1.8x that of running on a single machine.

If this holds true, Hy3 should end up running at over 30 tokens per second in a cluster. Performance like that would make Hy3 my go-to for local inference, and likely the first solution I'd recommend to clients who want to build a serious on-premise self-hosted LLM solution.

Add Mimo 2.5 and Stepfun 3.7 Flash models for multimodal needs, Deepseek V4 Flash for additional reliable coding and agentic capability (V4 Flash is still my main daily driver LLM), Qwen 3.6 35a3 MTP for fast sub-agent coding tasks, plus Gemma 4 26a4 QAT for fast multimodal sub-agent tasks (vision, audio, video), and that collection should form an extremely capable and performant group of digital brains to handle any challenge.

I'm very excited by the current state of all these models, which push the limits of what can be accomplished with relatively affordable local GPU hardware. Together with improved harnesses, we're pushing past what was possible with frontier models less than a year ago. Hy3, Deepseek V4 Flash, Mimo 2.5, and Stepfun 3.7 are near frontier, and Qwen 3.6 & Gemma 4 turn even small consumer GPUs into useful LLM machines - these are exciting times!

Hy3 is free to use on OpenRouter until July 21, 2026. Get to know it well, and build lots of software without spending a penny.

I made a Northwind database demo today:

http://1y1z.com:4832

It has all the typical bells and whistles:

• full CRUD editing of all fields in the database (create, read, update, delete)
• complete sort and filter functionality in all datatable displays (click column headers to sort)
• auth login to save changes to the database ('admin'/'admin')
• input/export CSV files to/from every table
• export order invoices to PDF
• and analytics page with charts
• responsive layout for mobile and desktop
• toggle UI between dark/light mode

Hy3 produced 100% perfect output for every software development prompt I gave it, and for all the agentic tasks I tried with it. It built everything very quickly without any hitches, and didn't choke for a moment. It's also done amazingly well with the suite of world knowledge tests I run on every new model. For a model in the several hundred billion parameter class, which can run on a single machine with 128GB RAM (DGX Spark, Strix Halo, and PCs with 4x 32GB GPUs), I haven't seen anything better yet.

So, this model is certainly going to be among my top 3 for local hosting.

My next tests will be with clustered DGX Sparks. From what I understand, vLLM already supports Hy3, with fast tensor parallelism. That means, using two Asus GX10s connected by the ConnectX ports, the expectation is 1.8x the speed of running on a single machine. If this holds true, Hy3 should end up running at over 30 tokens per second in a cluster, and performance like that would make Hy3 my go-to for local inference, and likely the first solution I'd recommend to clients who want to build a seriously powerful on-premise self-hosted LLM solution.

Add Mimo 2.5 and Stepfun 3.7 flash for multimodal needs, Deepseek V4 Flash for additional reliable coding and agentic capability (V4 Flash is still my main daily driver LLM), Qwen 3.6 35a3 MTP for fast sub-agent coding tasks, plus Gemma 4 26a4 QAT for fast multimodal sub-agent tasks (vision, audio, video), and that should form an extremely capable and performant collection of digital brains to handle any challenge.

I am very excited by the current state of all these models which push the limits of what can be accomplished with relatively affordable local GPU hardware. Together with improved harnesses, we're pushing past what was possible with frontier models less than a year ago. Hy3, Deepseek V4 Flash, Mimo 2.5, Stepfun 3.7, plus the Qwen 3.6 and Gemma 4 offerings, which turn even small consumer GPUs into useful LLM machines - these are exciting times!

HY3 is free to use on OpenRouter until July 21, 2026. Get to know it well, and built things without spending a penny. I made the Northwind database demo today, with all the normal bells and whistles:

http://1y1z.com:4832

• full CRUD editing of all fields in the database (create, read, update, delete)
• auth login to save changes to the database (admin/admin)
• complete sort and filter functionality in all datatable displays (click column headers to sort)
• export order invoices to PDF
• input/export CSV to/from every table
• analytics page with charts
• responsive layout for mobile and desktop
• toggle UI between dark/light mode

Hy3 produced 100% perfect output for every software development prompt I gave it, and all the agentic tasks I tried with it. It built everything very quickly without any hitches, and didn't choke for a moment.

This model is certainly going to be among my top 3 for local hosting.

My next tests will be with clustered DGX Sparks. From what I understand, vLLM already supports Hy3, with tensor parallelism. Using two Asus GX10s connected by the ConnectX ports, the expectation is 1.8x the speed of running on a single machine. If this holds true, Hy3 should end up running at over 30 tokens per second in the cluster, and performance like that would make Hy3 my go-to for local inference, and likely the first solution I'd recommend to clients who want to build an on-premise self-hosted LLM solution.

Add Mimo 2.5 and Stepfun 3.7 flash for multimodal needs, Deepseek V4 Flash for more reliable coding and agentic capability, Qwen 3.6 35a3 MTP for fast sub-agents, plus Gemma 4 26a4 QAT for fast multimodal subagents, and that should be an extremely capable and performant collection of brains to handle any challenge.