Latest7 min read
Agent alpha: a coding agent that runs in your browser
A real file workspace in OPFS, seven tools, approval before every write, and not one byte of your code uploaded. The alpha is open.
Read it
WebGPU.aiBlog
What we measure while making models run in a browser tab: download sizes, GPU buffer limits, sampling presets, and the things that break.
2 min read
Qwen3.5 4B: almost Gemma, for 2.2GB less
Why a fourth Qwen entered the catalogue, where it sits on the Pareto frontier, and which model to pick for your VRAM.
2 min read
Gemma 4 E4B: the biggest, and why we left its sampling alone
5.2GB of weights and 2GB of GPU buffer for the best quality we can run in a tab. Plus an afternoon lost to lowering a temperature we should not have touched.
2 min read
Qwen3.5 2B: the one we use ourselves
1.6GB, a dedicated runtime, a KV cache kept between turns. The smallest model of the catalogue on which an agentic loop holds.
2 min read
LFM2.5 350M: 255MB, and it loads everywhere
The smallest model of our catalogue, for the machines that refuse the others. Near-greedy sampling, 256MB of GPU buffer, and an honest list of its limits.
2 min read
Qwen3.5 0.8B: the model for a browser you know nothing about
470MB of q4f16 weights, 512MB of GPU buffer, and code that compiles. Why Qwen3.5 0.8B opened our on-device catalogue.