MOSS-Image: OpenMOSS's 5B Text-to-Image Model for Readable Text
MOSS-Image is OpenMOSS's open 5B text-to-image model: 96.3% long-text coverage, #2 behind GPT Image 2 in blind tests, Apache-2.0. How to run it.
Sumusブログ
AIチームメイトと働くためのアイデア、ガイド、最新情報。
DeepSeek 1017: V4.1 Flash 1017 Benchmarks, Pricing & Changes
DeepSeek 1017 (V4.1-Flash-1017, Oct 17, 2026): Terminal-Bench 31.2→42.3, shorter outputs, same $0.15/$0.60 price, 1M context. Full benchmarks and how to use it.
Claude Duet: Anthropic's Proactive Agent, Explained (2026)
Claude Duet is Anthropic's proactive agent app on Opus 5.5: morning briefs, background tasks, approvals. Waitlist, price, access and how it compares.
Claude Fable 5.5 Release Date: Still Not Out (Oct 28) — Latest Signals
Fable 5.5 didn't launch on Oct 27 as X posts claimed. Checked Oct 28: Anthropic still lists Fable 5.1. What's leaked, expected $10/$50 price, and the 3 signals to watch.
Inception Mercury 3: Benchmarks, 1,068 tok/s Speed & $0.10/1M Price
Mercury 3 benchmarks vs GPT-6 Luna and Haiku 5.5: 4–8x faster at 1,068 tokens/s, 1M context, $0.10/$0.35 per 1M. Free 100M tokens, chat demo and API in 3 lines.