Claude Sonnet 5, released by Anthropic on June 30, 2026, delivers performance nearly identical to the flagship Opus 4.8 model on most tasks while costing roughly 40 to 60 percent less. The mid-tier model introduces adaptive reasoning that adjusts automatically based on task difficulty, a 1 million token context window, and a focus on autonomous, multi-step work. For teams running frequent AI tasks, this represents a significant shift in the cost-versus-capability calculation.

What Makes Sonnet 5 Different From Previous Generations?

Sonnet 5 marks the largest generation-over-generation leap in the Sonnet line to date. The model introduces three technical changes that affect how users interact with it. First, it adjusts reasoning depth automatically based on the task at hand, replacing the manual extended-thinking toggle that earlier models required. This means simple prompts stay fast while harder problems receive more deliberation without any configuration needed from the user.

Second, Sonnet 5 uses a new tokenizer, the same change Anthropic introduced with Opus 4.7. This means the same input can map to more tokens, roughly 1.0 to 1.35 times depending on content type, so real per-task costs can run higher than the listed price suggests. Third, the model is tuned specifically for agentic work: planning, tool use, coding, and knowledge tasks that run across many steps rather than a single response.

How Does Sonnet 5 Actually Perform Against Opus?

The benchmark results reveal a striking pattern. On Terminal-Bench 2.1, a test of agentic coding ability, Sonnet 5 does not just close the gap to Opus 4.8; it surpasses it, scoring 80.4% compared to Opus's 74.6%, a jump of more than 13 percentage points over its own predecessor. On GDPval-AA v2, a knowledge-work benchmark, Sonnet 5 edges Opus 4.8 by a slim margin. This marks the first time a Sonnet-class model has outscored the concurrent Opus flagship on any benchmark.

Opus 4.8 still maintains a clear lead on the hardest coding and reasoning tasks. However, for the broad middle of everyday coding, agentic workflows, and knowledge work, Sonnet 5 handles the job at a fraction of the cost. The practical implication is straightforward: if you have been running a heavier model because earlier Sonnets could not keep up, that assumption is worth revisiting.

How to Choose Between Sonnet 5 and Opus for Your Workload

• Everyday Coding and Agentic Tasks: Sonnet 5 is the default choice, delivering near-Opus performance at roughly 40 to 60 percent of the cost. It handles multi-step workflows, tool use, and autonomous execution without sacrificing speed or accuracy for most professional applications.
• Hardest Coding and Complex Reasoning: Opus 4.8 remains the recommended model when you need the last increment of accuracy on extremely difficult tasks. It holds a clear lead on the most challenging coding benchmarks and complex reasoning problems that demand maximum capability.
• Cybersecurity Work: Opus is still Anthropic's recommendation for cybersecurity tasks. Sonnet 5 was deliberately restricted from this domain, so teams handling sensitive security work should continue using the flagship model.
• High-Volume Runs: Sonnet 5 becomes the economic choice when cost and speed matter across frequent calls. Its pricing of $2 per million input tokens and $10 per million output tokens makes it viable for applications that run dozens or hundreds of times daily.
• Knowledge Work: Sonnet 5 now slightly leads Opus on knowledge-intensive tasks, making it the preferred option for research, summarization, and information synthesis at lower cost.

The pricing structure makes the decision clearer. Sonnet 5 costs $2 per million input tokens and $10 per million output tokens, with these prices made permanent in August 2026. Opus 4.8 costs $5 per million input tokens and $25 per million output tokens. For teams running the same workload repeatedly, Sonnet 5 can reduce monthly bills significantly while maintaining comparable quality.

What Changed in Anthropic's Model Lineup?

Sonnet 5 now sits in the middle of Anthropic's Claude family, positioned between the fast, low-cost Haiku models and the high-capability Opus and Mythos-class models. The current lineup includes Claude Fable 5 at the top tier for the hardest, longest-running agentic work; Claude Opus 5 for demanding coding and reasoning at mid-high price; Claude Sonnet 5 as the broad default for most professional workloads; and Claude Haiku 4.5 as the fastest and cheapest option for high-volume, simple tasks.

Early testers reported that Sonnet 5 finishes complex tasks where previous Sonnet models would stop partway through, and that it checks its own output without being asked. It is a drop-in upgrade for the previous generation, Sonnet 4.6, though teams migrating existing workloads should recount a sample of their prompts before assuming costs stay flat, given the new tokenizer's impact on token counts.

The release of Sonnet 5 reflects a broader shift in how AI companies are approaching capability tiers. Rather than forcing users to choose between speed and accuracy, Anthropic has built a model that adapts its reasoning depth automatically, delivering flagship-level performance on many tasks at mid-tier pricing. For companies that have defaulted to expensive, high-capability models out of caution, Sonnet 5 offers a practical reason to reconsider that choice.