What Changed at Anthropic This Week
Alongside the return of Claude Fable 5 on July 1, Anthropic made Claude Sonnet 5 the new default model across its platform. Sonnet 5 at extra-high effort now performs roughly in line with Opus 4.8 on select agentic benchmarks while costing meaningfully less per token.
Anthropic replaced its previous default with Claude Sonnet 5, which at extra-high effort matches Opus 4.8 on key agentic benchmarks at lower cost. B2B teams relying on Anthropic defaults for sales workflows, content generation, and account research should test Sonnet 5 against current prompts this week.
For B2B enterprise teams, this is a practical workflow and budget decision, not just a technical announcement. If your team accesses Claude through the API without specifying a model version, you are now on Sonnet 5.
What Does Sonnet 5 Mean for Enterprise Workflows?
If you have been specifying Opus 4.8 for quality-critical tasks, test whether Sonnet 5 at extra-high effort produces equivalent output for your specific use cases. Key enterprise workflow categories to test:
Sales and outbound:
- Account research summarization from long-context inputs
- Personalization drafting for high-value outreach sequences
- CRM note generation from call transcripts and meeting recordings
Marketing and demand gen:
- Blog and long-form content drafting at scale
- Event topic research and briefing document generation
- FAQ and GEO content creation for programmatic pages
Security and compliance:
- Vendor risk assessment summarization and analysis
- Policy document review and gap analysis
- CISO briefing materials and board reporting drafts
For most of these tasks, Sonnet 5 at extra-high effort is expected to deliver output quality comparable to Opus 4.8. The cost difference is significant at scale.
What Is the Cost Implication for B2B Teams?
Switching from Opus 4.8 to Sonnet 5 for applicable workloads could reduce Anthropic spend by 40 to 60 percent on those tasks. For teams running Claude at scale for enrichment, summarization, or content generation, this is a real budget line impact.
Redirect savings toward higher-volume ICP targeting, better event production, or more precise account research that improves conversion from AI-sourced contacts.
What Is the Current Anthropic Model Family Structure?
The model family is now fully tiered across three capability levels:
- Fable 5 (restored July 1): frontier model for complex reasoning, large-context analysis, and the most demanding agentic tasks
- Sonnet 5 (new default): performance-tier workhorse for most enterprise agentic workflows
- Haiku 4.5: fast and cheap for high-volume, lower-complexity tasks
Plan your AI vendor budget around this structure. Fable 5 for reasoning-heavy tasks. Sonnet 5 for the bulk of enterprise workflows. Haiku 4.5 for volume operations where response speed matters more than depth.
What Should B2B Enterprise Buyers Do Right Now?
- Test Sonnet 5 against your existing Opus 4.8 prompts before relying on the new default in production
- Identify workflows where Sonnet 5 matches quality and shift them immediately to reduce cost
- Retain Opus 4.8 only for tasks where quality differences are material to business outcomes
- Update AI vendor budget models to reflect the new tiered structure and potential savings on Sonnet 5-eligible tasks