Built-in artifact log

Public artifacts, with the numbers and the blockers.

These are the publishable units of the TinyGPT factory: model packages, routed candidates, benchmark reports, and process artifacts. Each entry carries measured evidence, competitive context, and the exact reason it is ready, blocked, or parked.

8tracked artifacts
6active/reportable
2parked but public

Current Release Surface

SQL factory POC

Current-best candidate

Qwen3-0.6B Routed SQL Specialist

A two-adapter routed SQL artifact: public schema-only SQL routes to a b-mc2 adapter; local SQLite execution routes to a synthetic execution adapter.

Public exact
0.531
T5-small baseline
0.484
Synthetic execution
0.860

Competition: TinyGPT routed SQL v1 vs T5-small local baseline

Next: Publish this as a report artifact first. Do not present it as a shipped SQL model until public execution eval and clean-output gates pass.

First specialist package

Release-ready metadata

Qwen3-4B File-Ops Distilled

A routed 4B file-operation specialist distilled from frontier/gold trajectories, with the breadth regression disclosed in the package.

File-ops hard gate
100%
Heldout file-ops
95%
Breadth after tuning
42.3%

Competition: TinyGPT Qwen3-4B file-ops specialist vs Stock Qwen3-4B

Next: Release as metadata/model-card first, or publish the fused weights only with routed-only warnings attached.

Process artifact

Report artifact

Factory Run Schema v1

The canonical target -> data -> post-training -> eval -> package -> report shape for TinyGPT runs.

Required files
8
Decisions
6
First-class outputs
5

Competition: TinyGPT factory schema vs Ad hoc model card only

Next: Turn the SQL routed result into the first website-native factory report that follows this schema.

Browser performance artifact

Report artifact

Browser WebGPU Training Speedup

The original browser TinyGPT track: hand-written WebGPU kernels beat WASM SIMD more as model width grows.

WebGPU speedup
12.1x
Small-width speedup
2.6x
Browser track
shipped

Competition: TinyGPT WebGPU vs TinyGPT WASM SIMD

Next: Keep as a public performance artifact and cross-link it from factory reports when browser-local training matters.

Browser memory artifact

Report artifact

Memory64 Browser Behemoth Allocation

A WebAssembly Memory64 build lifted the browser model allocation ceiling past the old 4GB tab limit.

Allocated params
473M
Allocation time
3.7s
Train step
82.2s

Competition: TinyGPT Memory64 build vs TinyGPT wasm32 build

Next: Keep this as a public technical artifact; do not make it active factory work unless a browser-run specialist needs it.

Mac runtime benchmark

Report artifact

Huge Preset Decode Throughput

The native Mac runtime reached high local decode throughput on the Huge preset, showing the serving path is viable for local eval loops.

Huge decode
696 tok/s
Mega pilot
293 tok/s
Warm TTFT p99
5.8ms

Competition: TinyGPT Huge preset vs TinyGPT Mega pilot

Next: Use this as the baseline expectation for future artifact performance tables.

Parked Public Evidence

Mac runtime artifact

Parked

ANE M8 Core ML Chain

A layer-chunked Core ML chain ran a Qwen3 28-block path on the Apple Neural Engine at about 17 tok/s.

Distribution artifact

Parked

4-bit Browser Gallery Models

The browser gallery ships fp16 and int4 variants so model downloads are smaller and cold-start is cheaper.