Claude Sonnet 5.5 Arrives as Faster, Cheaper Complement to Opus 5.5

Claude Sonnet 5.5 is a mid-tier language model released by Anthropic on Sept. 28, 2026, as the second model in the Claude 5.5 family, priced at $2 per million input tokens and $10 per million output tokens and positioned as a faster, lower-cost complement to Claude Opus 5.5.

What is Claude Sonnet 5.5 and where does it sit in Anthropic’s lineup?

Anthropic released Claude Sonnet 5.5 on Sept. 28, 2026, as the second model in the Claude 5.5 family. The company positions it as a faster, lower-cost complement to Claude Opus 5.5, aimed at well-scoped everyday tasks such as coding, agents, and knowledge work. Anthropic’s launch post describes the model as a clear upgrade over Claude Sonnet 5, with speed and cost improvements that make it a practical default for routine workloads.

Pricing for Sonnet 5.5 is unchanged from Sonnet 5: $2 per million input tokens and $10 per million output tokens. That is half the rate of Opus 5.5, according to the model documentation published on Anthropic’s platform. The pricing decision signals that Anthropic intends Sonnet 5.5 to absorb a large share of production traffic that currently runs on more expensive models.

Anthropic says Sonnet 5.5 runs more than 30 percent faster than Sonnet 5 and costs up to 30 percent less per task. The per-task savings come not from a lower token price but from efficiency: the model completes typical work with fewer tokens and fewer tool calls. For teams running high-volume agent workloads, that difference can be larger than a headline price cut.

Why does per-task cost matter more than token price?

Most enterprise AI budgets are set around total spend, not unit prices. A model that costs the same per token but finishes a task in half the steps reduces the bill without any procurement negotiation. Anthropic says Sonnet 5.5 achieves its per-task savings by producing fewer tokens and fewer tool calls, which is a meaningful distinction for teams running agentic workflows where each step adds latency and cost.

The same logic explains why Anthropic kept the per-token price identical to Sonnet 5. Buyers do not need to reprice the model; they need to measure how much work it can complete for that price. The launch post frames Sonnet 5.5 as a model that does most work for less, not as a discounted version of a larger model.

How does Sonnet 5.5 compare with Opus 5.5 and Sonnet 5?

Anthropic published Terminal-Bench 4.0 results for all three models, and the mid-tier model leads the family on that evaluation. Claude Sonnet 5.5 scored 70.6 percent, compared with 10.3 percent for Claude Sonnet 5 and 66.4 percent for Claude Opus 5.5. The result does not mean Sonnet 5.5 outperforms Opus 5.5 on every task, but it shows the gap between tiers has narrowed in agentic terminal work.

Model	Release date	Input price per million tokens	Output price per million tokens	Terminal-Bench 4.0 score
Claude Sonnet 5.5	Sept. 28, 2026	$2	$10	70.6%
Claude Opus 5.5	Earlier in Claude 5.5 family	2x Sonnet 5.5	2x Sonnet 5.5	66.4%
Claude Sonnet 5	Earlier	$2	$10	10.3%

The comparison table shows the structural change in Anthropic’s lineup. Sonnet 5.5 keeps the per-token price of Sonnet 5 while delivering benchmark results closer to, or above, the flagship. Opus 5.5 remains the higher-priced option for complex reasoning and long-horizon projects, but buyers now have a data point that suggests many agentic tasks can run on the cheaper tier.

What changed in performance and efficiency?

According to Anthropic, Sonnet 5.5 runs more than 30 percent faster than Sonnet 5. In practice, that speed matters for interactive coding assistants and for agent loops where each step waits on a model response. The efficiency gain also lowers cost per completed task, which Anthropic estimates at up to 30 percent for most work.

The efficiency story is about tool use as much as tokens. Anthropic says the per-task savings come from fewer tokens and tool calls, which is a direct benefit for agentic systems that chain together many steps. A model that can complete a workflow in fewer round trips reduces both API spend and latency.

What do the Terminal-Bench 4.0 results tell us about agentic work?

Terminal-Bench 4.0 is the evaluation Anthropic chose to highlight in the Sonnet 5.5 launch. The score of 70.6 percent represents a large jump from Sonnet 5’s 10.3 percent and exceeds Opus 5.5’s 66.4 percent. For developers, the benchmark is a signal that Sonnet 5.5 was optimized for terminal and tool-driven tasks rather than only chat completions.

The benchmark gap does not establish that Sonnet 5.5 is generally smarter than Opus 5.5. It suggests the mid-tier model was tuned for the kind of structured, terminal-based tasks that make up much of everyday agent work. Buyers should treat the number as one input to model selection, not a complete ranking of capability.

What do early customers such as Epic Games report?

Epic Games, the company behind Unreal Engine and Fortnite, tested Claude Sonnet 5.5 before launch. Daniel Vogel, Epic’s chief operating officer, said the model matched a higher-tier quality bar in early testing. Anthropic published Vogel’s assessment in its launch post.

In Epic’s early testing, Claude Sonnet 5.5 cleared the same quality bar you’d expect from a higher-tier model, holding up on a system design audit and a data flow review. The new model managed tens of thousands of lines of code for gameplay system architecture, kept responses snappy, handled multi-hour tasks, and delivered with less prescriptive prompting.

Vogel’s description points to the workloads Anthropic is targeting: gameplay system architecture, data flow review, and multi-hour tasks. He said the model handled tens of thousands of lines of code and required less prescriptive prompting. Those are the conditions enterprise teams care about when deciding whether a mid-tier model can replace a flagship in production.

What does the release mean for enterprise AI budgets and workflows?

For enterprises, Sonnet 5.5 adds a routing option between small, cheap models and the Opus flagship. The unchanged per-token price means teams are not paying more for the upgrade from Sonnet 5, while the per-task efficiency gains can reduce total spend. The bigger financial question is whether workloads currently assigned to Opus 5.5 can move to Sonnet 5.5 without a quality drop.

Anthropic’s own framing answers part of that question. The company says Sonnet 5.5 is designed for well-scoped tasks, which suggests complex, open-ended reasoning should stay on Opus. Epic’s testing suggests some high-value engineering work can move down without sacrificing quality.

Development teams evaluating Sonnet 5.5 should consider the following steps before changing model routing:

1. Measure per-task token consumption and tool-call counts on representative workloads; Anthropic’s 30 percent savings estimate is based on typical work and will vary by use case.
2. Run the same agentic tasks on Sonnet 5.5 and Opus 5.5 to establish a quality baseline, since benchmark results do not guarantee parity in proprietary environments.
3. Compare latency in production, where Anthropic says Sonnet 5.5 runs more than 30 percent faster than Sonnet 5.
4. Review the cyber safeguards and usage policies for autonomous use cases, because Sonnet 5.5 is the first Sonnet model to launch with Opus-comparable safeguards.
5. Track total cost per completed task, not token price, when forecasting budget impact.

Model routing decisions are becoming a standard part of AI procurement. Sonnet 5.5 gives finance teams a concrete lever: the same token price as Sonnet 5, half the token price of Opus 5.5, and lower per-task consumption. The combination should make Sonnet 5.5 an attractive default for production coding and agent systems.

What safety protections ship with Sonnet 5.5?

Anthropic says Claude Sonnet 5.5 is the first Sonnet model to launch with cyber safeguards comparable to those for Opus models. The distinction matters because Sonnet models are often used in high-volume settings where automated agents take actions. The safeguard profile may remove a barrier to adopting Sonnet 5.5 in agentic workflows that handle sensitive systems.

For security teams, the inclusion of cyber safeguards at the Sonnet tier is a notable shift. Anthropic says Sonnet 5.5 is the first in the Sonnet line to carry protections comparable to Opus models, which lowers a barrier for autonomous and security-sensitive deployments.

What comes next for the Claude 5.5 family?

Sonnet 5.5 is the second confirmed model in the Claude 5.5 family. Anthropic’s launch materials describe the model as a complement to Opus 5.5 without detailing what comes next. The pattern so far is a flagship Opus followed by a mid-tier Sonnet with flagship-adjacent benchmark results.

The bigger signal is the inversion on Terminal-Bench 4.0, where Sonnet 5.5 outscored Opus 5.5. If model families continue to specialize by workload, pricing will increasingly reflect task-specific value rather than a single capability ladder. That would make model selection more complex for buyers but also more cost-effective.

Frequently asked questions about Claude Sonnet 5.5

The following questions cover release timing, pricing, benchmark results, and safety details from Anthropic’s documentation and launch post.