dor4.net

← AI players

Anthropic

Claude Fable 5, Opus 4.8, Sonnet 5, Haiku 4.5, and tools for long-running agents.

Anthropic builds the Claude family of models and the Claude assistant. Developers can work directly with models through the Messages API or use Claude Managed Agents for longer-running, asynchronous tasks with a pre-built agent harness.

Current Claude line-up

The main models are organized by the kind of trade-off they make:

Fable 5, Opus 4.8, and Sonnet 5 accept up to a million tokens of context, while Haiku 4.5 uses a smaller context window. All four accept text and image input, produce text, support multilingual work, and include vision capabilities.

Anthropic also offers Claude Mythos 5 through a limited-access program for specialized defensive cybersecurity workflows. It is not part of normal self-service model selection.

Reasoning and tools

The larger Claude models use adaptive thinking, allowing the model to vary its reasoning effort with task difficulty. Applications can also set an effort level when they need a deliberate quality, latency, or cost trade-off.

Claude can call custom functions and server-side tools for web search, web fetch, code execution, memory, file handling, and computer use. Structured outputs and strict tool schemas help applications consume model responses reliably.

Platform shape

The Messages API gives teams direct control over prompts, tools, conversation state, and orchestration. Managed Agents add infrastructure for tasks that need to run for longer periods. Claude models are also distributed through major cloud platforms, giving organizations a choice of deployment and data-routing arrangements.

Anthropic treats each modern model ID as a pinned model snapshot rather than an evergreen pointer. Moving to a newer Claude release is therefore an explicit migration that should be evaluated on representative tasks.