Anthropic has five models, a new series about every eight weeks, and one model it won't release to the public.
When you look across the AI landscape, you see a lot of activity from all the labs. Anthropic, OpenAI, Google, Meta, DeepSeek. They're all in play and all doing a lot. Today we're looking at Anthropic: what they're doing, where they're digging in, and the models they have out now.
To understand Anthropic, you need to see where it fits in the landscape. Plot the current models from the big labs. Up is smarter, right is more expensive. Then pull out Anthropic's.
Anthropic has five models. One of them, Mythos, isn't available to the general public. More on why below. The other four climb the price and performance curve. Haiku sits at the bottom, fast and cheap. Sonnet is a really strong everyday model. Opus is the deep one, the best for ongoing coding and the top-scoring model out there right now. Fable is for the longest-running jobs, but it's also really expensive. As people say, it burns your tokens fast.
Then there's the release pace. Anthropic's newest family is 5.5. It started in September with Opus 5.5. Sonnet followed six days later, and Haiku came out in early October. Fable is still on 5.1, and a 5.5 release is expected soon. Go back a year and count: seven series in about 12 months, 15 model releases, a new series about every eight weeks.
In April, Anthropic built a model better than anything seen before at coding tasks. With it, they identified serious bugs in operating systems, web browsers and thousands of open source projects. Because of that, distribution of Mythos was limited and rolled out slowly, so it could be used to harden technical infrastructure ahead of even more powerful models to come. Fable is a sibling of Mythos. It has a lot of the same capabilities, with different safeguards around it.
Haiku 5.5 came out in early October. It's the small, fast model: quick answers, sorting and short summaries. It's also what a lot of apps run underneath when they need to make thousands of calls. Anthropic also recently introduced two pricing tiers based on the total context used.
Sonnet 5.5 came out in late September. It's the everyday model, great for writing, analysis and a lot of coding. It's moderately priced and fits most people and most tasks. It uses a lot of tokens, so it can cost more than some competitors, but it's a great point in the price-value chain for Anthropic's models.
Opus 5.5 is the deep, complex one. It's my daily driver when I'm coding or doing deep analysis. It's somewhat expensive, but incredibly capable. It's good with long-running tasks and complex problem solving. With 5.5, Anthropic fixed some of its tendency to overspeak and get flowery.
Fable 5.1 came out in early September, so it's on the 5.1 cycle, not 5.5. It's designed for long-running jobs with lots of tasks built in. It investigates, it checks, and answers can take a long time, but it can go really deep on a subject. That doesn't always mean it benchmarks the best. It means it can deliver in-depth, quality results where that's needed. There are rumors a Fable 5.5 will come out shortly to join the 5.5 series.
So do you need to switch between models? For most people, no. Most people's work tends to be one kind of task. If you're in the Anthropic family, you're probably using Sonnet or Opus as your daily driver. You may occasionally step down if you're doing a lot of work and don't want to use up your tokens, or step up to a higher-level model for a particularly tricky task.
That's Anthropic. New models are shipping every couple of months, on roughly an eight-week cycle, and Anthropic has done a great job of constantly pushing the state of the art on what its models can do and where they're focused.