Claude Sonnet 5 Launches — Closer to Opus, and Cheaper
Anthropic released Claude Sonnet 5 on June 30 (local time). It's the successor to Sonnet 4.6, and the headline is simple: performance moves closer to the higher tier, Opus 4.8, while the price stays flat or drops. If you're on the Free or Pro plan, there's nothing to configure — it's already switched to your default model.
Sonnet 5 is now the default for Free and Pro
On the Free and Pro plans, Sonnet 5 is now the default model. Max, Team, and Enterprise users can all select it too, and it's available in Claude Code and the API (Claude Platform).
The practical part is that there's no separate signup or regional gating — it just shows up in the interface you already use. In Claude Code it appears in the /model picker under the model ID claude-sonnet-5.
Introductory pricing runs through August 31
API pricing at the introductory rate:
- $2 per million input tokens
- $10 per million output tokens
That rate applies through August 31, after which it rises to $3 input / $15 output.
Even at the standard rate it sits well below Opus 4.8, and Anthropic is pitching Sonnet 5 as "a cheaper option for running agents." If you're doing cost-sensitive automation or high-volume calls, it's worth benchmarking against the introductory rate before the end of August.
Per the official docs, the context window is 1M tokens and max output per call is 128K tokens.
Benchmarks reportedly approach Opus 4.8
According to Anthropic's announcement, Sonnet 5 is clearly ahead of Sonnet 4.6 on reasoning, tool use, and coding.
| Benchmark | Sonnet 4.6 | Sonnet 5 | Opus 4.8 |
|---|---|---|---|
| SWE-bench Pro (coding) | 58.1% | 63.2% | 69.2% |
| Terminal-Bench 2.1 (terminal work) | 67.0% | 80.4% | — |
| Humanity's Last Exam (reasoning, with tools) | — | 57.4% | 57.9% |
On SWE-bench Pro it scores 63.2%, above Sonnet 4.6's 58.1% but still short of Opus 4.8 (69.2%). On Terminal-Bench 2.1 it jumps to 80.4% from Sonnet 4.6's 67.0%. And on Humanity's Last Exam, a broad reasoning test, it hits 57.4% when using tools — nearly level with Opus 4.8 (57.9%).
That said, these are all vendor-reported figures, so the real-world feel is something you'll have to judge by using it yourself.
The positioning: "the most agentic Sonnet yet"
Anthropic describes Sonnet 5 as "the most agentic Sonnet to date." The focus is on making plans and working autonomously over long stretches while using tools like a browser and terminal.
To sum up: Opus 4.8 still holds the top of the accuracy curve, and Sonnet 5 lands close to it while cutting the price. For anyone who wants to run tool-heavy, agentic work without the cost pressure, it looks like a meaningful release.
Sources: Anthropic — Introducing Claude Sonnet 5, Claude Platform Docs — Pricing
This post summarizes the official announcement. It is not sponsored by Anthropic in any form.
Original with full infographics and visual structure: https://jessinvestment.com/claude-sonnet-5-launches-closer-to-opus-and-cheaper/
Top comments (0)