Welcome back to LLM Decode 👋
AI models are getting better at more than answering prompts- they’re becoming capable of long-running work, coding, reasoning, and specialized tasks. Today’s launches from Anthropic and Google show where the industry is heading next: models designed to do the work, while becoming faster and cheaper to deploy.
Anthropic Just Split Frontier AI Into Two Paths
Anthropic has launched Claude Fable 5.1, its latest model for coding, knowledge work, and complex agentic tasks, alongside Claude Mythos 5.1, a more restricted version aimed at specialized scientific and cybersecurity work.
Fable 5.1 focuses on making autonomous workflows more capable and economical. Anthropic says optimized caching can reduce the cost of typical agentic workloads by up to 45%, while cache-read pricing has been cut substantially.
Both models share the same underlying foundation but operate with different safeguards. Fable 5.1 is generally available, while Mythos 5.1 remains restricted to vetted organizations working in sensitive areas.
Why It Matters
Agents are becoming more economical: Lower operating costs make long-running AI workflows more practical.
Specialization is accelerating: One foundation can support different levels of capability and access.
Coding is becoming an agent-first use case: AI is moving from generating snippets to handling larger software tasks.
Safety is becoming part of model design: Frontier capabilities increasingly come with different access and safeguard layers.
Introducing The First Agentic CRM
Get revenue agents, workflows, and automations across every stage of your motion. Access customer data in real time through Attio's web app, MCP, API, and SDK.
Then Ask Attio anything about your business and get instant answers.
It's the CRM that runs the work behind every win.
Google’s “Fast” AI Is Getting Seriously Capable
Google has introduced Gemini 3.8 Flash, calling it its most intelligent Flash model yet for reasoning, coding, software engineering, and agentic workflows. It is designed to deliver stronger performance without the cost and latency traditionally associated with larger frontier models.
The model launches at an introductory price of $0.75 per million input tokens and $3.75 per million output tokens, matching Gemini 3.7 Flash's introductory pricing through the end of 2026. Google says the model can spend additional tokens and perform iterative tool calls on complex tasks to improve results.
Google also launched Gemini 3.8 Flash Cyber, a specialized cybersecurity model available through its Fairwind Program to trusted defenders. It focuses on vulnerability discovery and automated patching rather than offensive security capabilities.
Why It Matters
Small doesn't mean simple anymore: Faster models are increasingly capable of sophisticated reasoning and coding.
AI economics are changing: Lower-cost intelligence makes agentic applications easier to scale.
Specialized models are becoming strategic: Cybersecurity is emerging as a major domain for AI agents.
Watch token usage, not just token price: More reasoning can improve performance while increasing the total cost of a task.
4 AI Tools & Community Workflows
Agent Cost Audit: Before deploying an AI agent, measure not just model pricing but total tokens, tool calls, retries, context size, and task duration. The cheapest model per token isn't always the cheapest model per completed task.
Model Routing: Use a fast, lower-cost model for routine tasks and reserve expensive frontier models for complex reasoning. This can make AI workflows significantly more scalable.
Long-Horizon Coding: Instead of asking AI to generate individual code snippets, experiment with workflows where the model plans, writes, tests, reviews, and iterates across an entire task.
Specialized AI Stack: Don't assume one general-purpose model should handle everything. Watch the rise of domain-specific models for coding, cybersecurity, research, finance, and other professional workflows.
Ready to level up your AI skills?
Explore Our Courses →That’s it for today.
The AI space doesn’t slow down - and neither should your thinking.
See you in the next drop.