PulseAugur
EN
LIVE 08:23:35

Phancy releases open weights for Phai-A1 coding and agent model

Phancy, previously known as 4Paradigm, has released the open weights for Phai-A1, a model designed for coding and agent tasks. This model is based on the Qwen3.6-35B-A3B mixture-of-experts architecture. The training process utilized Huawei Ascend NPUs, and the model's checkpoint is available on Hugging Face and ModelScope. AI

IMPACT Provides an open-source option for coding and agent tasks, potentially fostering further development in these areas.

RANK_REASON Release of open weights for a model based on a mixture-of-experts architecture. [lever_c_demoted from research: ic=1 ai=1.0]

Read on Mastodon — mastodon.social →

AI-generated summary · Google Gemini · from 1 sources. How we write summaries →

Phancy releases open weights for Phai-A1 coding and agent model

How we ranked this

Signal score
18 / 100
Composite score across the factors below. Higher = stronger signal that this story matters right now.
Newsworthiness bucket
Tool
Release of open weights for a model based on a mixture-of-experts architecture. [lever_c_demoted from research: ic=1 ai=1.0]
Source corroboration
Single-source cluster
Only one publisher covered this so far. Single-source stories can still rank when the publisher is high-authority, but they lack cross-source corroboration.
Topics
model release
Editorial topic classification. Feeds into how the story surfaces on /topic/<slug> hub pages and into the per-entity coverage mix.
AI-industry relevance
High
Clearly on-topic for AI-industry coverage.
Story freshness
Breaking (< 6h)
Fresh story with cross-source coverage still developing. Ranking may shift as more sources report.

Full methodology in our editorial standards.

COVERAGE [1]

  1. Mastodon — mastodon.social TIER_1 English(EN) · [email protected] ·

    Phancy, formerly 4Paradigm, has published open weights for Phai-A1, a coding and agent model built on the Qwen3.6-35B-A3B mixture-of-experts base. Its fine-tuni

    Phancy, formerly 4Paradigm, has published open weights for Phai-A1, a coding and agent model built on the Qwen3.6-35B-A3B mixture-of-experts base. Its fine-tuning and reinforcement learning ran on Huawei Ascend NPUs, and the Apache 2.0 checkpoint spans about 70GB on Hugging Face …