Stories by steveharing1

ZAYA1-8B matches DeepSeek-R1 on math with less than 1B active parameters

Ling 2.6 (Flash and 1T): Efficient Open Models Competing on Agentic Benchmarks

Remodex: Control Codex from Your iPhone

AI Coding Models You Can Run Locally on Consumer Hardware

Open Design: Use Your Coding Agent as a Design Engine

Granite 4.1: IBM's 8B Model Matching 32B MoE

Laguna XS.2: A Coding Model Built for Closed Environments, Now Open Weights

Text-to-CAD: Generate 3D models with coding agents (open source)

Xiaomi releases MiMo-v2.5 Family weights with strong coding and agent benchmarks

Open CoDesign: Open-source, local-first alternative to Claude Design and v0

UniGenDet: Training image generators and detectors together

DeepSeek-V4: Making 1M token context efficient

Prax: An agent runtime that learns from past mistakes and fixes code in a loop

OpenMythos: A looped transformer take on how Claude Mythos might work

Anthropic's relationship with the Trump administration seems to be thawing

Alibaba open-sources Qwen3.6-35B-A3B, a 35B MoE model with 3B active parameters

Moss-TTS-Nano: Real-Time Voice AI on CPU (Open-Source TTS Stack)

Gemopus: A Gemma fine-tune that prioritizes stability over long chain-of-thought

An AI that researches the web before generating images

Multica: Assign issues to coding agents and track them like teammates

1 2 3