Anthropic centers Opus 4.8 as Claude family's high‑performance model while Sonnet and Haiku balance cost and speed

Anthropic now markets Claude Opus 4.8 as its top choice for complex research and coding, with Sonnet 4.6 positioned as the default cost‑aware model and Haiku 4.

Anthropic has repositioned models within its Claude family around a clear performance hierarchy: Opus 4.8 as the high‑performance option for the most complex tasks, Sonnet 4.6 as the all‑around default, and Haiku 4.5 as the fast, low‑cost choice. That positioning matters for teams deciding which model to run for research, agentic workflows, or production coding because it ties specific cost and latency tradeoffs to concrete work types.

Claude refers both to a set of large language models and to the chatbot interface built on them. The lineup now explicitly differentiates Haiku, Sonnet and Opus by their tradeoffs among speed, capability and cost, reflecting Anthropic’s shift from a conversational assistant toward a tool teams hand projects to and monitor as work progresses.

Anthropic lists per‑token pricing for each model: Sonnet 4.6 is priced at $3 per million input tokens and $15 per million output tokens; Opus 4.8 costs $5 per million input tokens and $25 per million output tokens; and Haiku 4.5 is $1 per million input tokens and $5 per million output tokens. The company also says Opus 4.8 represents a step up over Opus 4.7 across domains, including improved honesty about its capabilities.

Access tiers and usage limits are explicit. Free plans include limited access to Sonnet and Haiku; the Claude Pro subscription at $20 per month unlocks Opus plus features such as Claude Code, Cowork and Research. Power users can pay $100 per month for the Max 5x plan to multiply usage limits by five, or $200 per month to raise limits by 20×. Models are also available through app integrations and automation platforms for teams wiring AI into workflows.

The market context frames Anthropic’s choices: the company was founded in 2021 by former OpenAI researchers and trains Claude using a safety‑focused constitution it cites as a core development principle. Both enterprises and smaller teams are adopting Claude as a practical productivity tool, with Opus favored for agentic and research tasks and Sonnet used as the cost‑aware default for routine workloads.

For builders the tactical implications are straightforward: choose Haiku where latency and token cost dominate (customer service and high‑volume data processing), use Sonnet for balanced performance and token economics, and reserve Opus for accuracy‑sensitive research or complex engineering that can justify higher token spend. Teams should benchmark real task performance with tools such as AutomationBench and factor subscription limits into production deployment planning.