AI news

Claude received a significant upgrade — Anthropic is testing two new features (https://claude.com/resources/articles/dashboards-and-motion) that give the neural network the ability to create animations and interactive dashboards directly in the chat. Claude Dashboards — acts as an analytics expert. The user describes their project, and the AI automatically generates a dashboard that updates in re…

Anthropic has introduced Claude Haiku 5.5 — a model with maximum operating speed and the lowest cost in the new series. Key features: — 39% success rate in solving complex programming tasks, whereas in the previous version this figure was zero. — The cost of token processing decreased by 10 times. — In computer control tasks, a result of 72% was achieved, compared to 16% for Haiku 4.5. — In gener…

DeepSeek replaced Claude Code. The company has released an open-source desktop application. Features: • interacts with files, source code, and the terminal directly on your device • consists of plugins — you can add any features missing by default • searches the internet, generates text, and writes code • works fast, reliably, and without errors • supports Windows and macOS The main advantage is …

A new version of the neural network without filters has been released. The developers have fully figured out GLM-5.3 and removed almost all restrictions. What the model represents: • 753 billion parameters and a wide context • responds quickly and handles long, complex tasks • generates texts, answers, and instructions with virtually no blocks • easy to install, without unnecessary complications …

Google has completely surpassed Astra and Opus 5.5 by unexpectedly releasing Gemini 4 Argon. This model significantly outperforms the flagships from OpenAI and Anthropic on most tests: — The maximum output token limit reaches 1 million, although sometimes the neural network processes data more slowly. — It was designed specifically for complex and lengthy tasks in programming, business, and cyber…

OpenAI has once again taken the lead by unveiling the fresh GPT-6.1 Sol model. It has the same capabilities as Claude Opus 5.5, while being offered at a lower price. The competition in the field of artificial intelligence has once again turned into a battle of giants: one releases a more advanced version, another lowers the price, and a third is already preparing its response.

A new version of Claude Sonnet 5.5 has been released — a potential competitor to GPT-6 Sol — available in the app and on the website. The release surpasses GPT-6 Sol in efficiency when performing long-running tasks, nearly catches up to the leading Opus 5.5 model, and has significantly improved in programming. At the same time, operating speed has increased by more than 30% compared to Sonnet 5, …

Joke: GPT-6 Astra has a "turbo mode" — just mention that Claude handled it better. After that, the system seems to recharge: it starts functioning faster and takes on tasks it previously ignored. A useful tip for those who know how to effectively inspire artificial intelligence.

Specialists from NVIDIA have introduced Nemotron 3 Diarization — an advanced system for speech recognition. It is capable of analyzing audio tracks with multiple speakers, even if they are arguing, and accurately determining who is speaking and at what moment. The model supports dialogues involving up to eight people, which is the best result compared to analogues. At the same time, Nemotron 3 Di…

Guides for GPT-6 Sol and Opus 5.5 are already available online — the developers have released official materials with recommendations for using the new models. In short: GPT is great for writing texts and code in everyday tasks, while Opus is better at handling complex problems, graphics, agents, and large-scale programming. 🕤 For GPT-6 Sol and Luna (https://developers.openai.com/api/docs/guides…

Opus 5.5 has been released — Anthropic has announced its latest advanced neural network. According to the company, the new model achieves results comparable to Fable 5.1 on most tasks, while its cost is 20% lower than that of Opus 5. Opus 5.5 can already be tried via the link (https://claude.ai/new).

Anthropic has released Claude Opus 5.5. The model has closed in on Fable 5.1 and become cheaper Anthropic has introduced Claude Opus 5.5, the first model in the Claude 5.5 family. The company positions it as a more affordable alternative to Fable 5.1 for programming and complex intellectual work. According to Anthropic's estimate, a typical task costs 40% less than when using Opus 5, and response…

Grok 4.7 has been released. This is the most advanced version of Grok to date. What's changed: • a significant step forward compared to Grok 4.6 in all aspects • improved long-context handling • increased reliability when working with code and texts • most importantly — the cost has not changed compared to Grok 4.6 Try it here. (https://grok.com/)

Now you can run a huge model with 35 billion parameters directly on an iPhone — the open-source Edge0 engine has been introduced. It enables working with local neural networks based on the MoE architecture without creating a large load on RAM — at peak moments, only 1 to 2.5 GB is used. A trick is used for optimization: instead of RAM, the device's flash memory is utilized. Models with 8 billion …

GPT-6 uses 70% fewer tokens — specialists at Artificial Analysis tested the latest model and identified its key advantage. 🕤Scores 67 points in programming — comparable to Fable 5 and Claude Opus 5, while the model's operating cost is three times lower. 🕤Consumes up to 70% fewer tokens compared to GPT-5.6 Sol. 🕤Hallucination rate dropped from 92% to 51%, with accuracy improving by 4%. 🕤More e…

Claude takes first place once again. A new version, Fable 5.1, has been released, and based on test results, it significantly outperforms the previous Fable 5. In some benchmarks, quality has nearly doubled, while in others there is steady and consistent growth. Check it out for yourself on the website (https://claude.ai/new).

Qwen3.8-27B — this new Chinese AI is now available for free, with no filters or limits whatsoever. Previously, Qwen3.8-27B outperformed Claude Opus 4.6, and now the developers have removed all restrictions. • Built-in reasoning mode • Context supports up to 262,000 tokens • Handles text and images • No censorship and no restrictions Use it via the link (https://huggingface.co/huihui-ai/Huihui-Qwe…

SpaceXAI has released a new language model, Grok 4.6. In terms of intelligence, Grok 4.6 is comparable to GPT-5.6 Sol from OpenAI and Claude Fable 5 from Anthropic. In the Artificial Analysis Intelligence Index ranking, the new model scored 61 points. This matches the performance of GPT-5.6 Sol and trails Fable 5 Max by just one point.

OpenAI plans to remove limits on text-based conversations in ChatGPT for users on the free tier. These changes will roll out alongside the launch of GPT-5.6 Luna. However, restrictions on image generation and file uploads will remain unchanged.

Alibaba has unveiled its new flagship neural network, Qwen 3.8 Max, with 2.4 trillion parameters, which Chinese experts claim surpasses GPT-5.6 Sol and Fable. Key facts: • The AI can independently carry out long-running tasks for up to 16 days without human involvement; • In a demo, the model spent 12 hours autonomously designing a chip; • In benchmarks, Alibaba Qwen 3.8 Max outperformed GPT-5.6 …

Model Kimi K3 has appeared on Hugging Face. This Chinese giant has 2.8 trillion parameters, a context length of one million tokens, and a price significantly lower than comparable products from OpenAI and Anthropic. The model is already successfully competing with leading American developments. China is once again releasing an AI that stirs up a bit of excitement in the market. You can try it her…

Chinese developers have unveiled GLM 5.2 — a new open-source model already being hailed as one of the most powerful AI breakthroughs of the year. Here's what the creators and early users are saying: • The model delivers strong performance in tasks involving logical reasoning and programming; • In some benchmarks, it's being compared to leading Western counterparts; • Operating costs are significa…

⚡️Official release of GPT-5.6 — OpenAI has announced three new models positioned as serious competitors to Claude Fable 5. • Sol — surpasses Fable 5 in power, will be available to all paid subscribers within the next 24 hours. • Terra — similar in performance to Fable 5, will be offered for free. • Luna — also free to use.

Xiaomi has unveiled MiMo Code, which is already positioning itself as a competitor to Claude Code — the new AI model outperforms Anthropic's product on several benchmarks. What's promised: • the ability to work with context of any length — the neural network can handle projects containing millions of lines of code; • data retention when analyzing large repositories; • model training directly duri…

Editors, grab IMBU for video creation — the updated Aleph lets you change objects, backgrounds, and environments in just one click. The key advantage is that you only need to edit a single frame, and the model will automatically apply those changes across the entire video. Also added: — preview changes before rendering — videos up to 30 seconds long — 1080p resolution — more reliable scene proces…

⚡️OpenAI has announced an updated image generator — GPT-Image 2. This system can create realistic software interfaces, game scenes, advertising materials, "photographs" of newspapers, and even handwritten text — visually, the result resembles a real screenshot or camera photo. The creators note a significant reduction in artifacts and errors during generation. Access is now open to everyone. Join…

Google has made Veo 3.1 free for all users—it's an excellent tool for video creation, and the reason Altman shut down Sora. • Still considered one of the best video generators on the market. • Images come out with maximum detail. • Supports a variety of visual styles—from photorealism to cartoon animation. • Lighting, shadows, and composition are executed without any distortions. • Handles the Ru…