Every new model release and AI tool worth knowing about, translated out of jargon for beginning coders, hobbyists and anyone curious what changed this week.
AGENTS.md is a shared instruction file for AI coding agents. Claude Code 2.1.277 now reads it as a fallback, so one file works across every major coding tool.
Claude Cowork and chat merge into one interface. You no longer pick a mode first: Claude routes tasks, runs them in the background, and ships Docs and Slides.
Gemini 3.8 Live is Google's speech-to-speech AI that reasons and calls tools mid-conversation. Extended Thinking tops the Speech to Speech Quality Index at 82.6.
IFM's K2 Horizon 7B open-weights model matches models four times its size on coding and math. What beginners should know before trying it.
Ten consumer RTX 3090 GPUs now run DeepSeek V4 Flash Vision Exp, a 285B multimodal model, at 60+ tokens per second with FP4 and speculative decoding.
OpenAI now logs 3.1 agent-workdays per human workday, up from below 1.0 before June 2026, as coding agents write and run code on their own.
At $10 and $50 per million tokens, GPT-6 Astra dominates agentic work but ties GPT-5.6 Sol on general intelligence. Here is when to switch.
Claude Fable 5.1 is Anthropic's newest AI model with a 52.6 percent score on Terminal-Bench-Science 0.1, more than double Fable 5. Its five reasoning effort levels let beginners control cost and quality.
Gemini 3.8 Flash is Google's newest mid-tier AI model, released September 2, 2026, with better reasoning and coding than 3.7 Flash at the same introductory price. A security variant called Flash Cyber is gated to vetted defenders.
Community builders pushed Qwen3.8-27B to 2,000 tok/s prefill and 132 tok/s decode on a single RTX 3090. Here is what that speed means for local AI.
Gemini Omni 1.1 Flash is a Google DeepMind model for generating and editing video by conversation. You can extend clips to 40 seconds, control camera moves with keyframes, preview cheaply in 360p, and upscale to 4K.
GPT-5.6 in Kiro brings OpenAI models to the spec-driven coding tool with a major price drop: Luna falls to a 0.1x credit multiplier and Terra to 1.0x, making frontier AI coding agents cheap enough for hobbyists.
Qwen 3.8 27B is a free, open-weights AI model that runs on your laptop. After one week and 2,000 community posts, here is what testers found.
Mojo open source under Apache 2.0: the AI hardware programming language from Modular is now free to use and modify. Here is what changed.
Gemini 3.7 Flash is Google's new fast AI model for coding and agents. It scores 43.6 percent on FrontierCode, up from 34.4, at half the old price.
Only Free and Go users see ChatGPT ads. The sponsored placements launched in the U.S. in February 2026 and reached nine countries by August.
Qwen3.8 open weights give you 2.4 trillion parameters with 95 billion active per token, Alibaba's largest. Here is what beginners should know.
Meta's 30B open-weights Muse Glimmer fits on one RTX 3090 and reaches 280 tokens per second, built for local agentic workflows.
Claude Code auto mode runs commands without per-step permission. It becomes the default August 14, 2026 for Pro, Max, and Team plans.
OpenAI sharpened GPT-5.6 Sol for paying users and opened Luna to free users with unlimited text chats. Here is what beginners should know about the update.
Meta's Muse Code coding agent is co-trained with Muse Spark 1.2 for whole-project work; the contributor tier costs $0.10 per million input tokens.
Scoring 50 on the Intelligence Index, the DeepSeek V4 Flash 0731 open-weights update matches March 2026 frontier models. What it changes for you.
With 284B parameters and 13B active, DeepSeek V4 Flash went live on July 31, 2026, matching frontier coding benchmarks at a fraction of the cost.
OpenAI's GPT-5.6 price cut drops Luna to $0.20 per million input tokens and Terra by 20 percent, enabled by GPT-5.6 Sol optimizing its own inference stack. Here is what beginners should do.
This AI agent guide maps the shift from chat to agents. ChatGPT Work and Claude Cowork now lead for real work, while Gemini has dropped off the list.
At 2.8 trillion parameters, the Kimi K3 open weights released July 26, 2026 are the largest AI model download yet. What beginners can do with them.
Thinking Machines Lab's first open-weights model, Inkling, has 975B parameters, 41B active and an Apache 2.0 license, and targets fine-tuning over benchmarks.
Moonshot AI's 2.8 trillion parameter open-weight Kimi K3 matches top US closed models on key benchmarks, with weights due July 27.
Storing each parameter as a single bit lets Bonsai 27B fit 27 billion parameters into 3.9 GB and run on an iPhone. Here is what 1-bit quantization costs.
GPT-5.6 is OpenAI's newest model family in three sizes: Luna, Terra, and Sol. It claims big efficiency gains for long-running agent tasks at a fraction of competitor costs.
OpenAI's GPT-Live voice model listens and speaks at once, handing hard questions to GPT-5.5 mid-conversation without breaking the flow.
MTPLX v2 uses multi-token prediction to run local AI on Apple Silicon Macs up to 2.24x faster. Here is what beginners need to know.
Tencent's Hy3 open-weights model has 295 billion parameters but only 21 billion active per token, rivalling larger models and cutting hallucination to 5.4 percent.
MCP, the Model Context Protocol, is an open standard that lets AI models discover and call tools at runtime instead of you hardcoding an API for each one.
LongCat-2.0 is a 1.6 trillion parameter AI model from Meituan that activates only 48 billion parameters per token. Its weights are now open under the MIT license.
Early users report 20 tokens per second on a 26B model with GenieX, Qualcomm's runtime for running LLMs locally on Snapdragon laptops.
AI for dummies is the plain-language corner of Data Today. It covers new model releases and AI tools for people who are curious, learning to code, or building their first project, and who find the launch posts full of jargon they have no reason to know yet.
The Opinion Desk reads the same primary sources as the rest of the newsroom: the announcements from OpenAI, Anthropic, Google DeepMind, Meta, Mistral, and Microsoft, plus the Hugging Face blog and developer communities. It then writes an explainer that answers four questions: what was released, what it is good at, what it costs, and whether a beginner should try it.
Explainers here are opinionated by design. They say which tool to start with and which to skip, and they are labelled as explainers so you know that a recommendation is a judgement call. The facts behind the recommendation are still sourced and linked: benchmark numbers, prices, and context limits come from the vendor's own documentation or from public leaderboards.
What you will find here:
Every piece is reviewed by the responsible editor before it runs. The deeper technical coverage of the same releases lives in the AI section of the main site. The filters above separate new releases from beginner basics.
From the Data Today newsroom
These guides are part of Data Today. Read the wider data & AI coverage on the main site.
Go to Data Today