Kimi K3 review: Moonshot AI’s architecture, benchmarks, open weights, and Fable 5 comparison

3 August 2026
kimi k3

Kimi K3 review: Moonshot AI’s architecture, benchmarks, open weights, and Fable 5 comparison

Moonshot AI has released Kimi K3, the first open model in the 3-trillion-parameter class, with a 1-million-token context window. It is already available in the Kimi chatbot and through the…

Bonsai 27B: 1-Bit Weights Put a 27B-Parameter Model on a Smartphone for the First Time

15 July 2026
Bonsia 27B 1-bit

Bonsai 27B: 1-Bit Weights Put a 27B-Parameter Model on a Smartphone for the First Time

PrismML, a startup founded by Caltech researchers, has announced Bonsai 27B — binary and ternary versions of the Qwen3.6-27B model that retain 90–95% of the original model’s quality while compressing…

Claude Sonnet 5: A Strong Agentic Upgrade, but No Clear Opus Replacement

1 July 2026
claude sonnet 5 new

Claude Sonnet 5: A Strong Agentic Upgrade, but No Clear Opus Replacement

Anthropic has introduced Claude Sonnet 5, a new model in the Claude family that is also available to users on the free tier. It is designed for agentic tasks, programming,…

VibeThinker: 3B model reasons and codes at the level of flagship models

16 June 2026
https://neurohive.io/ru/ii-v-marketinge/pochemu-socseti-blokirujut-67-multiakkaunterov-v-pervye-3-dnya-analiz-500-otchetov/

VibeThinker: 3B model reasons and codes at the level of flagship models

Sina Weibo AI published VibeThinker-3B — a compact language model with just 3 billion parameters that matches flagship models DeepSeek V3.2 (671B), GLM-5 (744B), and Gemini 3 Pro on verifiable…

SenseNova-U1: NEO-unify multimodal architecture works directly with pixels without VAE

14 May 2026
SenseNova-U1 Unifying Multimodal model

SenseNova-U1: NEO-unify multimodal architecture works directly with pixels without VAE

SenseNova introduced a new multimodal architecture, SenseNova-U1, which combines image understanding, generation, and editing inside a single transformer without a separate visual encoder or variational autoencoder. This approach removes the…

OpenGame: AI Agent Generates Full Browser Games from Text Description

22 April 2026
gameengine

OpenGame: AI Agent Generates Full Browser Games from Text Description

A team of researchers from CUHK MMLab published OpenGame — the first agentic framework for creating browser-based 2D games from natural language descriptions. The project is fully open: the framework…

ChatGPT Images 2.0: OpenAI Launches Image Generation Model With Reasoning, 2K Resolution, and Multilingual Text

22 April 2026
gpt-images-2

ChatGPT Images 2.0: OpenAI Launches Image Generation Model With Reasoning, 2K Resolution, and Multilingual Text

April 21, 2026, OpenAI released ChatGPT Images 2.0 powered by the gpt-image-2 model. According to LM Arena, the new model immediately took first place across all image generation categories with…

InCoder-32B-Thinking: Open-Source Code Generation Model for Microcontrollers, GPU Kernel Optimization, and RTL Design

7 April 2026
Overview of InCoder-32B-Thinking

InCoder-32B-Thinking: Open-Source Code Generation Model for Microcontrollers, GPU Kernel Optimization, and RTL Design

A research team from Beihang University, Shanghai Jiao Tong University, the University of Manchester, and IQuest Research has published InCoder-32B-Thinking — a language model with an extended chain-of-thought reasoning for…

VBVR: 2 Million Videos for Reasoning Training — an Open Dataset That Changes the Rules

26 February 2026
VBVR dataset

VBVR: 2 Million Videos for Reasoning Training — an Open Dataset That Changes the Rules

A team of more than 50 researchers from around the world — from CMU, Oxford and other universities — has published Very Big Video Reasoning (VBVR), a massive dataset for…

GLM-5: Top-1 Open-Weight Model for Code and Text Generation, Competing with Claude and GPT on Agentic Tasks

19 February 2026

GLM-5: Top-1 Open-Weight Model for Code and Text Generation, Competing with Claude and GPT on Agentic Tasks

Zhipu AI and Tsinghua University have published a GLM-5 technical report — currently the top-performing open-weight language model by benchmarks: first place among open-weight models on Artificial Analysis and top-1…

Baichuan-M3: An Open Medical Model That Conducts Consultations Like a Real Doctor and Outperforms GPT-5.2 on Benchmarks

10 February 2026
Baichuan-M3

Baichuan-M3: An Open Medical Model That Conducts Consultations Like a Real Doctor and Outperforms GPT-5.2 on Benchmarks

A research team from the Chinese company Baichuan has introduced Baichuan-M3 — an open medical language model that, instead of the traditional question-and-answer mode, conducts a full clinical dialogue, actively…

Claude Sonnet 4.5 Leads on Comprehensive Backend Benchmark, Outperforming in Both Code and Environment Configuration

22 January 2026
abc-bench-pipeline-workflow

Claude Sonnet 4.5 Leads on Comprehensive Backend Benchmark, Outperforming in Both Code and Environment Configuration

A team of researchers from Fudan University and Shanghai Qiji Zhifeng Co. introduced ABC-Bench — the first benchmark that tests the ability of AI agents to solve full-fledged backend development…

Multiplex Thinking: Sampling 3 Tokens Instead of 1 Increases Olympiad Problem-Solving Accuracy from 40% to 55%

22 January 2026
multiplex thinking

Multiplex Thinking: Sampling 3 Tokens Instead of 1 Increases Olympiad Problem-Solving Accuracy from 40% to 55%

Researchers from the University of Pennsylvania and Microsoft Research introduced Multiplex Thinking — a new reasoning method for large language models. The idea is to generate not one token at…

P1: First Open-Source Model to Win Gold at the International Physics Olympiad

30 November 2025

P1: First Open-Source Model to Win Gold at the International Physics Olympiad

P1-235B-A22B from Shanghai AI Laboratory became the first open-source model to win a gold medal at the latest International Physics Olympiad IPhO 2025, scoring 21.2 out of 30 points and…

Which AI Can Play a Villain: Comparing Alignment Algorithms Across 17 ModelsRetry

13 November 2025

Which AI Can Play a Villain: Comparing Alignment Algorithms Across 17 ModelsRetry

Researchers from Tencent Multimodal Department and Sun Yat-Sen University published a study on how large language models handle role-playing. It turns out that AI models perform mediocrely at role-playing: even…

DeepEyesV2: Multimodal Model Learns to Use Tools to Solve Complex TasksRetry

12 November 2025
deepeyesv2-illustration

DeepEyesV2: Multimodal Model Learns to Use Tools to Solve Complex TasksRetry

Researchers from Xiaohongshu introduced DeepEyesV2 — an agentic multimodal model based on Qwen2.5-VL-7B that can not only understand text and images but also actively use external tools: execute Python code…

From Millions Spent on “Thank You” to Efficient Inference: Boilerplate Detection in a Single Token

31 October 2025
Detecting Boilerplate Responses LLM

From Millions Spent on “Thank You” to Efficient Inference: Boilerplate Detection in a Single Token

Researchers from JFrog published a study demonstrating a method for early detection of boilerplate responses in large language models after generating just a single token. The method enables computational cost…

QeRL: Training 32B Models on Single H100 vs Three GPUs, Beating LoRA in Accuracy

16 October 2025
QeRL rainforcement learning quantization training speedup

QeRL: Training 32B Models on Single H100 vs Three GPUs, Beating LoRA in Accuracy

QeRL is a framework for training language models using reinforcement learning that simultaneously reduces GPU requirements and surpasses traditional LoRA and QLoRA methods in accuracy. On the Qwen2.5-7B-Instruct model, QeRL…

Mini-o3: A Multimodal 7B Model Outperformed GPT-4o in Visual Search With 30-Step Reasoning Chains

10 September 2025
Observation mini-o3

Mini-o3: A Multimodal 7B Model Outperformed GPT-4o in Visual Search With 30-Step Reasoning Chains

Researchers from ByteDance and the University of Hong Kong introduced Mini-o3 — a multimodal model that performs deep multi-step reasoning to solve complex visual search tasks. Mini-o3 achieves SOTA results…

NVIDIA Nemotron Nano 2: Reasoning and Code Generation Model Outperforms Qwen3-8B on Benchmarks and Supports 128k Context

20 August 2025

NVIDIA Nemotron Nano 2: Reasoning and Code Generation Model Outperforms Qwen3-8B on Benchmarks and Supports 128k Context

A team of NVIDIA researchers presented Nemotron-Nano-9B-v2 — a hybrid Mamba-Transformer language model that generates responses 6 times faster than Qwen3-8B on reasoning tasks while exceeding it in accuracy. The…

3D-R1: Open Source Reasoning Model for 3D Scenes Outperforms State-of-the-Art Methods by 10% on 3D Benchmarks

6 August 2025
3D-R1 model

3D-R1: Open Source Reasoning Model for 3D Scenes Outperforms State-of-the-Art Methods by 10% on 3D Benchmarks

Researchers from Shanghai University of Engineering Science and Peking University presented 3D-R1 — a new foundation model that significantly improves reasoning capabilities in three-dimensional vision-language models (VLM). The model demonstrates an average performance…