Nicheloom visitors: 12,433 all-time · 7,595 in the last 7 days

The opportunity tracker for new startups.

Today: 970 launchs · Niche of the day: Windows desktop utilities and tools · See full report →

Phocinae-Largha-150M-v1

斑海豹 — Phocinae-Largha-150M-v1: a 144.3M bilingual decision model. Tiny model. Big decisions. 小海豹,大决断。

Get picks like this daily. The day's top launches, AI/tech news, and a weekly opportunity spotlight — straight to your inbox.

This is 1 of 188 launches in niche digital tools and utilities — see how it stacks up on momentum and crowding →

51 other launches read as similar to this one →

Details

External ID
1409518007
Source
GITHUB
Company
—
Product
Phocinae-Largha-150M-v1
Website domain
github.io
Launched
Oct. 8, 2026
Cohort
—
Upvotes
60
Upvotes percentile
0.76
Tags
bilingual, calibration, cpu-inference, decision-making, decision-model, local-inference, modernbert, option-order-invariance, small-model, structured-decisions, text-classification, typed-decisions, zero-output-tokens

Enrichment

Niche
niche digital tools and utilities
Vertical
Horizontal
Function
Model & infra
Audience
Developer
AI stance
AI-native
Project type
Hobby / open-source project
Normalized one-liner
compact bilingual decision model for developers
Manually corrected
False

Could you build this?

No This is a custom-trained, calibrated 144M-parameter bilingual transformer decision model requiring deep machine learning research, custom tokenization/encoder architectures, and curated bilingual training datasets.

What it would actually take: Developing this requires PyTorch, custom architecture modification of an encoder (mmBERT-small), curated supervised dataset pipelines for non-autoregressive classification/decision tasks, and GPU training clusters. It requires specialized NLP/ML expertise in calibration, loss functions, and inference optimization (ONNX/TensorRT/GGUF).

Competitors

Other products that read as similar to this one — 51 launches clear the similarity bar, closest 8 shown.

Attention rank: #19 of 52 (itself plus its competitors, highest first — normalized so YC and Product Hunt are compared fairly).

Launched 283 days after the earliest competitor.

Other launches for this product