Research

Haru-4B model: natural interruptions without a TTS chain

On 13 November 2026, Higurashi released Haru-4B, a 5.3-billion-parameter voice model that listens and speaks every 133 ms. It can replace a separate text-to-speech stage for live, open-ended conversation,...

Viggle Turbo’s Sigma Schedules Explained

Viggle Turbo v0.3’s correct default is six Euler steps using raw sigmas . Use that exact list for normal generation and editing. The faster...

MiniMax M3.1 Flash shows strong agent scores but a reliability gap

MiniMax M3.1 Flash scored 43 on Artificial Analysis’s Intelligence Index in October 2026. The independent score placed it fifth among 173 models in its...

AI jailbreaks on GitHub need a red-team intake gate

FuturPulse analysis · 23 October 2026 Alejandros17487/ai-jailbreaks describes 35 paste-ready prompts for more than 25 model targets and instruction surfaces. Teams should not paste them...

Qwen Flash Next on an RTX 4090: what the 100 tokens/second claim means

FuturPulse analysis · 6 October 2026 A single RTX 4090 has not been benchmarked at 100 tokens per second for Qwen3.8-Flash-Next. As of 6 October...

LLM Benchmark Leaderboard: How to Read Rankings Without Buying the Wrong Model

FuturPulse analysis · 5 October 2026 On 5 October 2026, Scale’s leaderboard puts GPT-6 Astra at 60.60 on Humanity’s Last Exam Diamond, while Artificial Analysis...

Best AI for Research: How to Pick a Model for Files, Forecasting, and Million-Token Sources

FuturPulse analysis For academic research, Consensus is the strongest starting point because it searches a the public record of more than 200 million scholarly papers...

Discover OAT: Revolutionizing Robotics with Action Tokenization

Key TakeawaysOAT, developed by researchers from Harvard and Stanford, enhances robotics with scalable actions. It converts continuous robot movements into discrete tokens using a transformer...

Great Refactor Initiative: AI Transforms Critical Code to Rust

Key TakeawaysThe initiative aims to convert vulnerable C and C++ code into the more secure Rust language, improving software safety. AI tools can significantly automate...

Agentic RL Training Insights for GPT-OSS: A Practical Overview

Agentic RL training — Key TakeawaysAgentic RL optimizes entire decision-making processes through interaction with environments. Training in agentic RL involves iterative closed loops for policy...

AI Co-Scientist in Medical Knowledge Synthesis: A Proof of Concept

AI Co-Scientist — Key TakeawaysAI Co-Scientist is designed to assist in knowledge synthesis specifically in medical contexts. The paper presents a proof of...