Claude now has four current models side by side, and the newest, Claude Fable 5.2, costs twice as much per token as Claude Opus 5.5. So the practical question for anyone working in Claude Code or on the API is not “which model is best?” but “which model fits this task, at what cost?” Anthropic’s own documentation gives a fairly clear answer. Start with the job you need done.
Where to start: Opus 5.5 at medium effort
Anthropic’s model overview says that if you are unsure, you should start with Claude Opus 5.5 for most workloads. It describes Opus 5.5 as built for long-running agentic coding and knowledge work. “Agentic” here means the model works through a task in many steps, using tools such as a terminal or files, rather than giving one answer. On the API, its default effort level is medium.
Before switching models, try a different effort level
Effort is a setting that trades intelligence for speed and cost inside a single model: higher effort lets the model think longer before it answers. Anthropic’s guide on choosing a model says tuning effort is often a better lever than switching models. In practice, if an Opus result falls short, raising the effort is the smaller and cheaper first step, and changing the model comes after that.
When Fable 5.2 is worth the higher price
Anthropic calls Fable 5.2 its most capable model open to all customers, built for demanding reasoning and long-horizon agentic work, meaning tasks that run over hours with many decisions along the way. Its listed price is $8 per million input tokens and $40 per million output tokens, against $4 and $20 for Opus 5.5. A token is a small chunk of text, and providers bill by the million. Fable 5.2 is cheaper than Fable 5.1, which costs $10 and $50. Anthropic’s advice is to move to Fable 5.2 when your tests on Opus at higher effort, such as xhigh or max, still do not meet your requirements. Its examples include agent sessions that run for hours and multistep research carried through to a finished document.
When Sonnet or Haiku is the better fit
Going the other way, Claude Sonnet 5.5 costs $2 per million input tokens and $10 per million output tokens, half the Opus price, and Anthropic positions it as the best combination of speed and intelligence for everyday coding, data analysis and agent work. Claude Haiku 5.5 starts at $0.10 and $0.50 and is meant for high-volume, latency-sensitive jobs such as classification, extraction and routing, for example sorting incoming customer messages.
Splitting one job across several models
You do not have to pick one model for a whole project. Anthropic describes two common patterns: a cheaper executor model that escalates hard decisions to a stronger advisor, and an orchestrator that hands bulk work to lower-cost workers, so most tokens are billed at the lower rate. In Claude Code, those workers are subagents, separate sessions that each handle one part of a job and report back. Anthropic also notes that with Fable 5.2, how much work goes to subagents depends on your setup, and subagents can account for a large part of what a session costs.
Our assessment
The documentation is consistent: Opus is the starting point, effort is the first dial, and Fable is the escalation for work where quality clearly outweighs a doubled token price. Our inference from that is simple. Before moving a whole project to Fable, run the same two or three real tasks on Opus at a higher effort level and on Fable, and compare both the result and the bill. If the difference does not show up in your own work, the cheaper model is the right one for now.
