Z.ai confirmed that GLM-5.3-Flash (320 billion parameters) runs on 100 000 Chinese AI chips instead of Nvidia GPUs. Released on 26 August 2026 under the MIT license, the model achieves performance close to GLM-5.3 at roughly one-seventh of the price.
New: The model was predicted on 20 August under the codename 'Ox Alpha', officially released on 26 August 2026; It is being deployed on 100 000 Chinese AI chips instead of Nvidia; Temporary price of 0.075 USD per million input tokens (until 9 September), then 0.15 USD; output tokens 0.25 USD/million; Focused on long contexts, agentic tasks and code synthesis
The Mogan AI team released the 149M-parameter Turkish encoder model MoganBert-TR, trained on 237.3 billion tokens using the CLM-to-MLM procedure; it achieves 78.41 points on TrGLUE, the best result among Turkish ModernBERT models. The accompanying smaller embedding model achieves 99.5 % of the performance of the teacher.
IBM has released Granite 4.2 open-weight models in three sizes (3B, 8B, 30B parameters) with a native context window of 128 000 tokens. The 8B and 30B variants have undergone agentic training for working with the terminal, web search and external tools.
AWS described a system for automated metadata harmonization built on Amazon Bedrock: it combines LLMs, embeddings and traditional NLP to correct and align schemas, but final approval of changes remains with the user.
Importance: ▮ monitoringGitHubNew models✓ official
GitHub has launched a public preview of an integration that brings GitHub Copilot agent mode to Slack. The team assigns a task by mentioning @GitHub, the agent works asynchronously in a dedicated code channel, and the PR can be made subject to approval before merging.
Generalist AI introduced GEN-1.5, a model that, according to the company, learns a new task from a single 3 to 12-second demonstration without further training. Without training, it achieves a success rate of 59 %, rising to 83 % with ten training steps on five minutes of data. The results have not been independently verified.
NVIDIA has made the open model Nemotron 3.5 Lightning (30B parameters, 3B active, MoE architecture) available in Amazon SageMaker JumpStart. According to the vendor, it offers up to 4× higher throughput and 30 % faster task completion for high-volume agent workflows.
The Intern team published a description of Intern-S2-Preview-397B (397 billion parameters) on arXiv for multimodal scientific reasoning and long agentic tasks. According to the authors, it achieves competitive to leading results on scientific and agentic benchmarks; a separate Memory Decoder module enables specialization…
Suno Studio 2.0 adds MIDI editing, a chat interface for creating instruments and effects, and stem separation; Premier subscribers get unlimited 32-bit/48 kHz export. Meanwhile, a German court ruled that Suno violates copyright law.
New: Stem separation is included in the Studio 2.0 features; Unlimited multitrack export at 32-bit/48kHz for Premier subscribers; The chat interface lets users communicate with the DAW as if it were a bandmate to create instruments and vocals; A German court ruled that Suno violates German copyright law
The vision-language model LFM2.5-VL-3B (3.1 billion parameters) has been released for edge devices. According to the developers, it improves screen reading, working with multiple images and function calling, has open weights and runs directly on the device.
The model LFM2.5-VL-3B is available on Hugging Face and in Playground. It supports function calling from text and image inputs. According to Liquid AI, it improves screen understanding, object localization and handling multiple images.
Google Research and Google DeepMind presented the AMIE research system, which conducts clinical video consultations in real time. In a study with simulated patients, it achieved expert-level performance in both diagnosis and communication, according to the company; it remains a research system with no deployment plan.
NVIDIA has published a technical description and benchmark results for the Nemotron 3.5 Super (120B/12B active) and Lightning (30B/3B active) models. The Lightning model itself was released by the company as open source under the OpenMDW 1.1 license back on 11 August, together with the NeMo Switchyard routing tool.
New: Lightning is a MoE hybrid Mamba-Transformer model with 3B active parameters out of 30B total; Super is a 120B model with 12B active parameters, multimodal, intended for code engineering; Both models include long-context training and native speculative decoding; The post-training pipeline uses RLVR, RLHF, and for Super, multi-teacher on-policy distillation (MOPD); Super achieves 51.78% on SWE-bench Pro and has 4.6× fewer active parameters than Nemotron 3 Ultra
NVIDIA has made the open-source model Nemotron 3.5 Lightning (30B parameters, 3B active per token, MoE) available in Ollama for local AI agents. It supports a context of up to 1M tokens, runs on NVIDIA GPUs as well as Apple silicon and can be further trained for custom tasks.
Meta Platforms launched the open-source model Muse Glimmer, a smaller alternative to models from competitors, designed to run directly on users’ devices with consumer graphics cards, without needing a cloud connection.
To accompany the release of the open model Muse Glimmer (30B, Apache 2.0), Meta published an extensive essay by Mark Zuckerberg defending distillation and its open strategy against OpenAI and Anthropic; it also promised to open the weights of Muse Spark 1.2.
New: Mark Zuckerberg published an essay (6000+ words) on AI philosophy, governance and the strategy of Meta; The announcement includes Muse Code, an AI agent for the terminal; Meta strategically positions open-weight models as an alternative to proprietary approaches from OpenAI and Anthropic; The essay criticizes lobbying by competitors and mentions competition from Chinese open-weight models; The new strategy of Meta focuses on differentiation through an open-source approach
According to sources familiar with the matter, ByteDance is training an AI model with up to 10 quintillion parameters — three times as many as Kimi K3 from Moonshot. It is in an early stage of pretraining; the estimated size of Mythos 5 from Anthropic is around 8 quintillion parameters.
Mistral has released the open-weight model Shieldstral (3 billion parameters) for classifying the safety of text and images. According to the company, it achieves an F1 score of 84.9 % on text (on par with the seven-times-larger GPT-OSS-Safeguard-20B) and 83.8 % on images. Rules can be configured at runtime in natural language without training.
Importance: ▮ monitoringNVIDIANew models✓ official
NVIDIA released the Alpamayo 2 Super model for commercial use in robotaxis and autonomous vehicles. According to the company, it is intended to help handle rare and complex traffic situations beyond standard object detection and motion prediction.
Companies offer AI "thanabots" trained on the messages, photos and voices of deceased people for 199 to 50 000 USD. Microsoft has a patent for this from 2020, while HereAfter AI is currently shutting down its service.
Common tags
LLM 18OpenAI 17open-source 17Hugging Face 16NVIDIA 13API 12
Step 1 of 4 · How AI Radar works
One clear overview from dozens of sources
AI Radar monitors Czech and international sources every day, looking for changes that truly deserve attention.
MonitorsOfficial AI company blogs, specialist media, and research sources.
Selects and combinesFilters out information noise and combines articles about the same change into a single event.
Summarizes and explainsExplains significant events in English: what happened, why it matters and where the information comes from.
The result is a quick overview of what has actually changed in the AI world, rather than another stream of articles.
Use the CS/EN switch to read the same Radar in Czech or English. English content is published after its translation has been checked, so new and older items may appear later.
Everything you need to navigate the AI world
Today’s briefingAn overview of news from the past 24 hours and a main story when available. Beside them is Live · AI Flash, followed by research and links to other Radar sections. On mobile, the blocks appear one below another.
AI FlashAn ongoing feed of brief updates with an evidence status. Links lead to a Radar detail page when one is ready, otherwise to the original source. You can also find reset and outage histories here.
Practical applicationsWhat new tools and features can do, what you can try and what their actual impact could be.
Model selectionModel comparison by type of work, capabilities, price and speed.
Research and archiveA separate research overview, topic search and older events by date.
One event, everything that matters
Each row represents one event — not one article. At a glance, you can see its significance, credibility and main point.
Illustrative example, not a current news item.
Importance: ▮▮▮ majorOpenAIModels✓ 6
Agent mode is available to all paying users
Until now, the mode was available only on the highest plan; it is now available on all paid tiers without a waitlist.
▮▮▮ major · ▮▮ important · ▮ we're tracking = how significant the change is✓ 6 = six independent publishers, not the number of articles or feeds✓ official = a clear release, law or incident is substantiated by the relevant authority1 source = no independent confirmation yetbold = who is behind the changegray text = a brief summary of what happened
The detail page contains a fuller summary, its significance and original sources. Practical impact appears in the detail and the For individuals and For businesses views. An AI Flash item reaches the main selection only after it has been expanded and meets the publication rules.
The same news, two practical uses
We first summarize each event in the same way for everyone. Based on those same facts, we then explain what the change means for your own use and what it could mean for how a company operates.
For individualsWhat you can use or try, how the change can help you at work and what to watch out for.
For businessesWhat impact the change could have on processes, costs, risks and other business decisions.
Today’s briefing is the same for everyone. Pages
For individuals and For businesses
can be found in the main navigation — they select only events relevant to the given use case.
