Claude Code

Latest Claude Models

Last updated 2026-07-28

What's new

2026-07-28
  • Microsoft released Mage Flow, an open-source (free, editable code) image tool that generates and edits images, like changing backgrounds or poses, with impressive speed and quality.
  • Shot Plan, a new open-source AI, creates videos with precise cuts, transitions, and camera movements, following detailed instructions for consistent, high-quality results.
  • GLM 5.2, a text-only AI model, now includes vision capabilities, expanding its functionality.
  • Google's latest Gemini models offer incredibly fast performance, and Alibaba teased their massive, open-source Qwen 3.8 model.
2026-07-25
  • A new AI design tool called Kimi K3 is challenging the top-ranked Claude, offering similar quality at 30% lower cost.
  • Both tools were tested on creating slides, dashboards, and websites, judged on output, cost, and speed.
  • Kimi K3 can be accessed via an API key from openrouter.com (a website that connects different AI tools) and used within Claude's coding system.
  • For audio narration, the presenter used Fish Audio (a cheap, voice-cloning service with 83+ languages and emotion customization) instead of 11 Labs (a more expensive alternative).
2026-07-16
  • Anthropic, an AI company, released a controversial ad highlighting potential AI risks, like job loss, homelessness, and societal collapse, to position itself as a responsible industry leader.
  • Anthropic's CEO estimates a 25% chance AI could cause catastrophic outcomes, including massive job losses in white-collar sectors within 1-5 years.
  • Anthropic and other AI companies face criticism for using vast amounts of public data to train AI models, then restricting others from learning from their outputs.
  • Businesses are investing heavily in AI agents (specialized AI tools), with some spending $3,000 to $10,000 monthly for services that boost productivity and efficiency.
2026-07-07
  • Anthropic is releasing a new AI model called Fable, which is designed to be more powerful and versatile, like moving from a tutorial to an open-world game.
  • Fable can perform complex tasks by using tools, like finding specific Pokémon names, which shows its "capability overhang" or spiky intelligence.
  • Claude Code, a tool that lets you give the AI direct commands, has been improved by reducing its system prompt, making it more imaginative and less constrained.
  • A new feature called the "ask you a question" tool allows the AI to interact with you in a multiple-choice format, making communication easier.
2026-07-04
  • Claude Fable 5 (a new AI model) is now available, offering better reasoning and understanding, but it's expensive and has limited free usage until July 7th.
  • To get the best results, give it the "why" behind your task and tell it what not to do, like an intern who needs clear instructions.
  • HyperAgent (a tool built by the Airtable team) lets you create a team of AI agents with different roles to give you honest feedback on your ideas.
2026-07-01
  • Google DeepMind is already planning for Artificial Super Intelligence (ASI) (AI smarter than all humans combined), not just Artificial General Intelligence (AGI) (AI as smart as a typical human).
  • They predict that AGI could lead to ASI through scaling (bigger, better AI models) or algorithmic shifts (new AI architectures or training methods).
  • AI is advancing so fast that researchers are now writing papers with instructions for AI to summarize them, assuming AI will read them instead of humans.
  • The paper also discusses a theoretical "universal AI" (AIXI), the ultimate limit of AI intelligence, which we can approach but never truly reach.
2026-06-28
  • Claude Fable 5, a powerful AI model (a type of AI that understands and generates human-like text), is expected to return soon, with high odds (90%) of launching by July 31st, after being taken offline due to security concerns.
  • Anthropic, the company behind Claude, accused Alibaba of stealing AI capabilities without paying, highlighting ongoing AI security challenges.
  • OpenAI, another AI company, released GPT 5.5, a more conversational AI model, and unveiled Hal Pino, a custom AI chip for faster processing.
  • Google DeepMind, a major AI research company, is facing setbacks, with researchers leaving and new AI models performing worse than older versions.
2026-06-25
  • Anthropic launched Claude Tag, a team-focused AI tool (like a smart assistant) that operates through Slack, helping teams write and manage code, analyze data, and more, with all work visible in public channels.
  • Claude Tag can proactively surface stalled discussions, handle long-term tasks, and assign work, all while maintaining separate data access for different teams, and it's currently in beta for enterprise users.
  • Anthropic quietly reduced the thinking depth of Claude Code (another AI tool) by about 67%, leading to user complaints, and users found that the AI's reasoning text was replaced with a cryptographic signature.
  • Caphathi is a new marketplace where creators can sell their AI agent skills (like writing scripts or auditing code) as closed-source apps, allowing them to earn money while keeping their expertise private.
2026-06-22
  • Anthropic released Claude Fable 5 and Mythos 5, AI models (computer programs that can learn and make decisions) designed to be more capable and useful for developers and companies, but faced governance challenges within 72 hours.
  • These models use a safety system with classifiers (fast guards that check if requests or answers fit policy) to reduce risky behavior, but also raise questions about access, safeguards, and public accountability.
  • The models aim to shift from being helpful assistants to more persistent partners in knowledge work, but their adoption depends on trust in the surrounding rules and infrastructure.
  • The launch sparked debates about access, with identity, intent, and audit trails (records of who did what and when) becoming crucial for governance and safety.
2026-06-16
  • The US government ordered Anthropic (an AI company) to shut down its advanced AI models, Fable 5 and Mythos 5, for everyone outside the US, including its own foreign employees, due to national security concerns.
  • Fable 5, launched just days before the shutdown, was designed to bring advanced cybersecurity capabilities (like finding and fixing software flaws) to the public, but a jailbreak (a method to bypass safeguards) was demonstrated to government officials.
  • Anthropic had implemented strong safeguards and monitoring to prevent misuse, but the government's action was unprecedented and lacked transparency, leading to criticism from the tech community.
  • This isn't the first clash between Anthropic and the Trump administration, as they were previously labeled a supply chain risk (a designation usually for foreign adversaries) after refusing to let the military use their models for certain purposes.
2026-06-13
  • Enthropic released Claude Fable 5, a powerful AI model (a computer program that learns and makes decisions) for the public, which is a less capable version of Claude Mythos 5, initially restricted due to security concerns.
  • Fable 5 can be connected to various apps (like Gmail, Google Calendar, and Notion) through connectors (tools that link AI to other apps), allowing it to perform tasks like finding and contacting potential customers automatically.
  • Clay, a connector acting as a company directory, helps Fable 5 search for and gather information about people from millions of companies, which can be used to draft and send personalized emails directly from Claude.
  • The Claude desktop app, featuring a workspace called Claude Co-work, enables users to organize and manage these AI-assisted tasks within specific folders and files on their computer.

Key points

What it is

  • Claude is an AI system designed for real-world tasks, like building apps, creating presentations, or planning. It can handle complex, multi-step jobs on its own.
  • The latest versions (Opus 4.7, 4.8, Mythos, and Fable 5) focus on "agentic ability" (working independently) and can do more complex tasks over time.
  • These models offer "effort control" (adjusting for simpler or more complex requests) and "skills" (reusable workflows tailored to your work style).
  • Anthropic (the company behind Claude) plans to release even smarter models that can improve themselves.

How to use it

  • Update your Claude code environment (a tool for using Claude) and call the model directly by typing `claude fable` or selecting Fable 5 from the model menu.
  • Give Claude tasks like building a simple app or creating a diagram to see how it works. Break complex tasks into steps.
  • Use Claude Fable for general coding and Claude Mythos 5 (if you have access) for autonomous software engineering or life sciences research.
  • Treat Claude like a thought partner (an equal collaborator), not a step-by-step assistant. Share your full context and invite its reasoning.

Watch out for

  • Don't treat Claude as a simple task-follower. Present your goal and ask for strategic input instead of giving step-by-step instructions.
  • Understand the naming confusion. Claude Mythos is a raw frontier model with limited access, while Claude Fable 5 is a safer version for general use.
  • Avoid switching tools too quickly. Stick with one model long enough to learn its strengths, as the models evolve fast.

Tools named

  • Claude (Anthropic's AI system for real-world tasks), Claude Code (a tool for using Claude to build apps), Claude Design (a tool for structured workflows)

Lesson 1: What is Latest Claude Models and why it matters

The latest Claude models, such as Opus 4.7 and 4.8, are the newest versions of Anthropic’s AI system designed for serious, real-world work. Unlike older models that only performed well on tests, these models focus on agentic ability—meaning they can handle complex, multi-step tasks on their own. The length of tasks Claude can complete independently has been doubling every four to seven months, showing accelerating improvement.

This matters for AI development because Claude has become a tool for turning ideas into finished products. You can build an app, create a presentation, organize research, or plan a week without needing a full team. Anthropic plans to release an entirely new class of models with intelligence beyond Opus, which could allow future versions to be continuously improved by Claude itself.

The models also offer effort control, letting you adjust for simpler or more complex requests, giving you control over speed and cost. Skills (reusable workflows inside Claude) let you build a system around how you work—knowing your projects, voice, and workflows. For coding and serious work, people actually want to use Claude because it can be handed an underspecified problem and figure out how to solve it.

Sources

Lesson 2: How to use Latest Claude Models: step-by-step

To use the latest Claude models—Mythos, Fable, and Insane—you first need to understand the new naming. Claude Fable 5 is a "Mythos class" model, meaning it belongs to a tier that sits above Opus in capability. Claude Mythos 5 is the most powerful version, but it’s currently available only to a small subset of users. Most beginners will access Claude Fable 5 first, as it has been released for general use.

To start, update your Claude code environment. Open your terminal (a text-based interface for commands) and run the update for your Claude service. Once updated, you can call the model directly. For example, type `claude fable` or select Fable 5 from the model menu inside Claude Code or Claude Desktop. If you see a model called Mythos in your interface, it means you have access to the preview version—use it similarly.

A concrete step: inside Claude Code, instruct the model to build a simple app by saying “Create a to-do list app with a diagram of the structure.” Fable 5 will generate code and, if prompted, create a visual diagram. This helps you learn how Claude reasons. For complex tasks, break them into steps. Use Claude Fable for general coding; switch to Claude Mythos 5 for autonomous software engineering or life sciences research if you have access. The key difference: Mythos class models work autonomously longer and cost less than half the price of the earlier Mythos preview. Always test prompts using a benchmark tool to compare outputs yourself.

Sources

Lesson 3: Best practices and pitfalls

The newest Claude models—Mythos and Fable 5—represent a new tier above Opus, but beginners often trip on three key pitfalls. First, do not treat these models as simple task-followers. Anthropic’s tips for Fable 5 emphasize treating Claude like a thought partner (an equal collaborator), not a step-by-step assistant. The old habit of saying “do this, then this” wastes capability. Instead, present your goal and ask for strategic input.

Second, understand the naming confusion. Claude Mythos is the raw frontier model, available through limited early access programs like Project Glasswing. Claude Fable 5 is a “Mythos class model” — the same underlying power but with stronger safety guardrails for general public use. So if you access Claude normally, you are likely using Fable 5, not the pure Mythos. Do not expect Mythos-level freedom in Fable.

Third, avoid switching tools too quickly. Rumors and leaks cause many to abandon Claude just as they become comfortable. The models evolve fast: recent checkpoints like Claude Fable 5 and Claude Fruitcake EAP show Anthropic is releasing multiple variants. Stick with one model long enough to learn its strengths.

Best practice: treat Claude as a peer, not a tool. Share your full context and invite its reasoning. For complex projects, use Claude Code or Claude Design for structured workflows. And when you see a new name like Mythos or Fable, check whether it is a public release or a limited preview before changing your setup.

Sources
← Back to curriculum