> Canonical page: https://posttrainllm.com/artifacts

Six published models · six case studies

# Every public model, with the result and the failure mode.

Every PostTrainLLM model on Hugging Face now has a first-class case study, including routed wins, a sealed-eval rejection, a negative-transfer failure, a conversion-only release, and an unevaluated checkpoint. Hub request counters are not treated as users or adoption. The rest of the registry preserves benchmark reports and process artifacts.

 6 published model case studies

 17 tracked public artifacts

 2 parked but public

## Current Release Surface

Agent-behaviour experiment

 Report artifact

### [OffHours Context-Interference Benchmark](https://posttrainllm.com/artifacts/offhours-context-interference)

A paired office-work benchmark that separates unresolved personal context from interruption and raw context volume. Devin showed no mental-toll penalty, but reached a reproducible context-volume boundary.

**Clean accuracy**
 99.0%

**Mental-toll effect**
 not detected

**Volume boundary**
 8,000 words/day

 Competition: Clean workday vs Resolved family context

 Next: Freeze this report-only result. A local provenance-complete comparison, if ever desired, begins as a fresh experiment after the learning phase.

A failed model that won us an answer

 Report artifact

### [Needle 2: The 45M Capacity Boundary](https://posttrainllm.com/artifacts/needle2-tool-selection)

Catalog restriction and a four-arm successor factorial both failed, establishing that this 45M model class—not one missing recipe tweak—is the current tool-selection bottleneck.

**Tiny-overfit gates**
 4/4

**Best successor**
 26/94

**Destructive bypasses**
 10 vs 2

 Competition: Needle 2 full catalog vs Needle 2 oracle task catalog

 Next: Close the 45M recipe-search lane. Keep the complete factorial as a hands-on learning artifact; any 1.7B successor is a fresh experiment.

Reference-scored browser ASR

 Report artifact

### [Parakeet WGSL Browser ASR](https://posttrainllm.com/artifacts/parakeet-wgsl-browser-asr)

Parakeet v3 scored 0/82 word errors and ran 3.51x faster than native WhisperKit on the same eight LibriSpeech clips, while missing a deliberately aggressive 50x short-clip speed gate.

**Browser WER**
 0.0%

**Native WER**
 8.54%

**Paired latency**
 3.51x

 Competition: Parakeet TDT 0.6B v3 · browser vs WhisperKit Mac-native baseline

 Next: Closed as a quality and native-latency win with a formal speed-threshold miss. Use it to teach paired evaluation, fixed-cost amortization, and why frozen gates matter.

SQL factory POC

 Report-ready candidate

### [Qwen3-0.6B Routed SQL Specialist](https://posttrainllm.com/artifacts/qwen06-sql-routed-v1)

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: posttrainllm routed SQL v1 vs T5-small local baseline

 Next: Keep the published report artifact report-only. Two hygiene DPO retries proved reference anchoring cures the SimPO collapse and even lifts execution to 0.920, but output hygiene needs a generation-strength fix (stronger SFT or constrained SELECT-prefix decoding), not more preference tuning. Do not present it as a shipped SQL model until public execution and clean-output gates pass.

 Report only — no model to use not fully verified [Report card →](https://posttrainllm.com/report-cards/qwen06-sql-routed-v1)

49.5M on-device classifier

 Release-ready weights

### [Pace Intent Router v8](https://posttrainllm.com/artifacts/pace-intent-router-v8)

A 49.5M intent classifier that was extremely fast and strong on its source-matched synthetic holdout, then failed the first leakage-checked sealed distribution gate.

[Hugging Face model ↗](https://huggingface.co/posttrainllm/pace-intent-router-v8)

**Sealed accuracy**
 57.1%

**Frontier retained**
 57.1%

**Warm latency**
 3.8ms

 Competition: Pace Intent Router v8 vs Qwen3-4B-Instruct-2507 4-bit

 Next: Do not tune on sealed V1. Build a broader public-development corpus from the failure themes, train a successor, and judge it once on a newly generated sealed V2.

First specialist package

 Release-ready weights

### [Qwen3-4B File-Ops Distilled](https://posttrainllm.com/artifacts/qwen3-4b-file-ops-distilled)

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

[Hugging Face model ↗](https://huggingface.co/posttrainllm/qwen3-4b-file-ops-distilled)

**File-ops hard gate**
 100%

**Heldout file-ops**
 95%

**Breadth after tuning**
 42.3%

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

 Next: Keep routed-only warnings attached and add a consumer pull/load smoke before wiring this into any app.

 Shipped — routed only not fully verified [Report card →](https://posttrainllm.com/report-cards/qwen3-4b-file-ops-distilled)

Routed depth and safety win

 Release-ready weights

### [Qwen3-4B ReST Fused](https://posttrainllm.com/artifacts/qwen3-4b-rest-fused)

Fresh requalification confirms a perfect, safer, faster file-ops specialist while rejecting it as the general successor because breadth regressed.

[Hugging Face model ↗](https://huggingface.co/posttrainllm/qwen3-4b-rest-fused)

**File-ops hard gate**
 12/12

**Qualified breadth**
 25/45

**Depth wall speedup**
 2.42x

 Competition: posttrainllm Qwen3-4B ReST vs Stock Qwen3-4B

 Next: Keep this package behind the file-ops route and reject it as the general Pace successor.

 Shipped — routed only not fully verified [Report card →](https://posttrainllm.com/report-cards/qwen3-4b-rest-fused)

Negative-transfer case study

 Report artifact

### [Qwen3-4B Multibackend Distilled](https://posttrainllm.com/artifacts/qwen3-4b-multibackend-distilled)

A public failed attempt that retained 100% file-ops depth but drove the recorded breadth score down to 31%.

[Hugging Face model ↗](https://huggingface.co/posttrainllm/qwen3-4b-multibackend-distilled)

**File-ops depth**
 100%

**Breadth**
 31%

**Breadth delta**
 -28.6pp

 Competition: Multibackend-distilled 4B vs Stock Qwen3-4B

 Next: Keep the weights public as a failed comparison artifact. Do not spend compute revalidating them unless a new breadth-preserving recipe explicitly needs this checkpoint as its baseline.

Conversion case study

 Report artifact

### [VibeThinker-3B MLX Conversion](https://posttrainllm.com/artifacts/vibethinker-3b-mlx)

A public Apple-Silicon-friendly conversion of WeiboAI/VibeThinker-3B, preserved as a runtime artifact rather than claimed as a PostTrainLLM-trained model.

[Hugging Face model ↗](https://huggingface.co/posttrainllm/vibethinker-3b-mlx)

**Model class**
 3B

**Local GSM8K screen**
 40/40

**Training delta**
 None

 Competition: PostTrainLLM VibeThinker-3B MLX vs WeiboAI/VibeThinker-3B

 Next: Keep this as a conversion and runtime artifact. Add a pinned loader plus parity receipt only when a real Mac-local consumer justifies maintaining it.

Missing-evidence case study

 Blocked

### [VibeThinker-3B Agentic Distilled](https://posttrainllm.com/artifacts/vibethinker-3b-agentic-distilled)

A preserved agentic distillation checkpoint whose public weights outlived its evaluation evidence; no current win can be claimed.

[Hugging Face model ↗](https://huggingface.co/posttrainllm/vibethinker-3b-agentic-distilled)

**Public weights**
 Yes

**Current agentic eval**
 Missing

**Reasoning retention**
 Unknown

 Competition: VibeThinker-3B agentic distilled vs VibeThinker-3B MLX base

 Next: Leave the weights public but unpromoted. If this lineage becomes active again, start with an evidence-only baseline/candidate evaluation—not more training.

Artifact storage cleanup

 Report artifact

### [Hugging Face Specialist Model Archive v1](https://posttrainllm.com/artifacts/hf-specialist-model-archive-v1)

The local specialist model cache was promoted to Hugging Face or deleted when safely re-downloadable from upstream repos.

**posttrainllm HF repos**
 6

**Local model cache**
 cleared

**Storage policy**
 HF first

 Competition: Hugging Face artifact storage vs Local Mac cache

 Next: Use this as the storage index; use the six dedicated model case studies for quality and decision evidence.

Process artifact

 Report artifact

### [Factory Run Schema v1](https://posttrainllm.com/artifacts/factory-run-schema-v1)

The canonical target, data, post-training, evaluation, package, and report shape for PostTrainLLM runs.

**Required files**
 8

**Decisions**
 6

**First-class outputs**
 5

 Competition: posttrainllm factory schema vs Ad hoc model card only

 Next: Use the routed SQL report as the canonical example; make live commands emit the same complete contract automatically.

Browser performance artifact

 Report artifact

### [Browser WebGPU Training Speedup](https://posttrainllm.com/artifacts/browser-webgpu-speedup)

A frozen ABBA run on an Apple M5 Pro measured a 10.67x Large-preset WebGPU speedup over WASM with 4.72% paired final-loss drift and zero runtime errors.

**WebGPU speedup**
 10.67x

**Loss drift**
 4.72%

**Hardware gate**
 pass

 Competition: posttrainllm WebGPU vs posttrainllm WASM SIMD

 Next: Keep Large as the measured M5 Pro win; reproduce Small, Medium, and XL independently before promoting the historical curve.

Browser memory artifact

 Report artifact

### [Memory64 Browser Behemoth Allocation](https://posttrainllm.com/artifacts/memory64-browser-behemoth)

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: posttrainllm Memory64 build vs posttrainllm 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](https://posttrainllm.com/artifacts/huge-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: posttrainllm Huge preset vs posttrainllm 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](https://posttrainllm.com/artifacts/ane-m8-coreml-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](https://posttrainllm.com/artifacts/gallery-int4-browser-models)

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

 From package to decision [Use the local-model evaluation workflow →](https://posttrainllm.com/evaluate-local-llm)