Anthropic released Claude Sonnet 5 on June 30, 2026. It is the new default model for the Free and Pro plans, and it is the most agentic Sonnet model the company has shipped. The headline is simple: Sonnet 5 gets close to Opus 4.8 on the benchmarks that matter for real work, but it costs a fraction of the price.
I have spent the hours since launch reading the announcement, the system card, and the early partner feedback. Here is everything a developer needs to know before switching.
Quick specs
| Spec | Claude Sonnet 5 |
|---|---|
| API model string | claude-sonnet-5 |
| Codename | Fennec |
| Release date | June 30, 2026 |
| Context window | 1,000,000 tokens |
| Inputs | Text, image, file |
| Effort levels | low, medium, high, max, x-high |
| Input pricing (intro) | $2 / 1M tokens through Aug 31, 2026 |
| Output pricing (intro) | $10 / 1M tokens through Aug 31, 2026 |
| Input pricing (standard) | $3 / 1M tokens |
| Output pricing (standard) | $15 / 1M tokens |
| Availability | Claude API, Claude Code, Free and Pro default, Amazon Bedrock, Microsoft Foundry, Google Vertex |
What is Claude Sonnet 5?
Sonnet 5 is the latest model in Anthropic's mid-tier Sonnet line, the direct successor to Sonnet 4.6. For many developers the agentic era began with Sonnet-class models, since Sonnet 3.5, 3.6, and 3.7 were the first to show strong coding and tool use. Lately the biggest jumps came from the larger Opus models. Sonnet 5 narrows that gap.
Anthropic describes it as built to act, not just answer. It makes plans, drives browsers and terminals, checks its own output without being asked, and runs autonomously for long stretches. That kind of sustained work needed bigger and pricier models only a few months ago.
Benchmarks: how good is Sonnet 5?
The numbers confirm the positioning. Sonnet 5 is a clear improvement over Sonnet 4.6 and lands within striking distance of Opus 4.8.
| Benchmark | Sonnet 5 | Sonnet 4.6 | Opus 4.8 |
|---|---|---|---|
| SWE-bench Pro | 63.2% | lower | 69.2% |
| OSWorld (computer use) | 81.2% | 78.5% | higher |
| SWE-bench Verified | strong | 79.6% | 88.6% |
| GPQA-AAA v2 | slight edge over Opus 4.8 | lower | high |
A few takeaways:
- On SWE-bench Pro, Sonnet 5 reaches 63.2% versus Opus 4.8 at 69.2%. That is close for a model that costs less than half as much.
- On OSWorld, the computer-use benchmark, Sonnet 5 hits 81.2%, a real step up from Sonnet 4.6.
- It actually edges Opus 4.8 on GPQA-AAA v2, a graduate-level reasoning test, which shows how much capability Anthropic packed into the smaller tier.
- It does not beat Opus 4.8 overall. Opus 4.8 still wins on the hardest coding and agentic tasks, especially at higher reasoning effort.
For the full picture of where the flagship still leads, see our Sonnet 5 vs Opus 4.8 comparison and the generational jump from Sonnet 4.6.
Pricing and the tokenizer catch
Sonnet 5 launches with introductory pricing of $2 per million input tokens and $10 per million output tokens through August 31, 2026. After that it moves to $3 input and $15 output. For reference, Opus 4.8 is $5 input and $25 output.
There is one detail most launch coverage misses. Sonnet 5 uses an updated tokenizer, the same kind of change Anthropic introduced with Opus 4.7. The same text can map to roughly 1.0 to 1.35 times more tokens depending on content type. Anthropic set the introductory price so the move from Sonnet 4.6 is roughly cost-neutral, not a flat discount. We break down what this means for your real bill in Claude Sonnet 5 pricing explained.
Effort levels
Sonnet 5 exposes selectable reasoning effort: low, medium, high, max, and x-high (extra high). Higher effort means more tokens and more accuracy. At its maxed-out x-high setting, Sonnet 5 performs about in line with Opus 4.8 at a medium-to-high setting on OSWorld and the agentic search benchmark BrowseComp. The twist is that running Sonnet 5 at x-high can cost more than running Opus 4.8 at a comparable accuracy point, so effort tuning is now a core cost lever. Our effort levels guide covers when each setting is worth it.
Safety and the cyber angle
Anthropic's pre-deployment evaluations found Sonnet 5 is safer than Sonnet 4.6. It refuses malicious requests more reliably, resists prompt-injection hijacks better, and shows lower rates of hallucination and sycophancy.
On cybersecurity, Anthropic says it did not deliberately train Sonnet 5 on cyber tasks. In a test of building exploits for Firefox 147 vulnerabilities, Sonnet 5 never produced a full working exploit. Cyber safeguards ship on by default, but they are lighter than the strict ones attached to the banned Fable 5. Because the cyber risk is low, the odds of a government pull like the one that hit Fable 5 are low too. We dig into that in Will the US government ban Sonnet 5?.
Who is already using it
Anthropic's early-access partners describe a model that finishes jobs older Sonnets would abandon. Cursor reports agents that stay on plan and ship clean multi-step changes at an efficient cost. Lovable highlights clean refusals of unsafe requests. ClickHouse points to tighter reasoning steps and faster time to insight. Testers describe Sonnet 5 writing a reproducing test, fixing a bug, then stashing the fix to confirm the bug returns, all in one pass.
How to start using Sonnet 5
- In Claude Code, set it as your model. See our Sonnet 5 Claude Code setup.
- Through the API, use the model string
claude-sonnet-5. See how to use the Sonnet 5 API. - With Aider, follow the Sonnet 5 Aider setup.
- Through a multi-provider router, see the Sonnet 5 OpenRouter setup.
If you are coming from the flagship to save money, read migrate from Opus 4.8 to Sonnet 5 first.
Frequently asked questions
Is Claude Sonnet 5 better than Opus 4.8?
No, not overall. Opus 4.8 still leads on the hardest coding and agentic tasks. Sonnet 5 gets close on most benchmarks and edges Opus 4.8 on GPQA-AAA v2, all at a much lower price.
How much does Claude Sonnet 5 cost?
Introductory pricing is $2 per million input tokens and $10 per million output tokens through August 31, 2026, then $3 and $15. Opus 4.8 is $5 and $25.
What is the context window?
One million tokens, large enough to load a full codebase in a single prompt.
Is Sonnet 5 free?
Yes. It is the default model on the Free and Pro plans, and it is also available to Max, Team, and Enterprise users.
What does the codename Fennec mean?
Fennec is Anthropic's internal codename for the model. It has no functional meaning for users.
Will the US government ban Sonnet 5 like Fable 5?
It is unlikely. Sonnet 5 has much weaker cyber capabilities than the Mythos-class Fable 5, which is the reason Fable 5 was restricted.
The bottom line
Sonnet 5 is the new value default for agentic coding. It will not replace Opus 4.8 for the very hardest problems, but for the bulk of day-to-day engineering work it delivers most of the quality at a fraction of the cost. With introductory pricing live through August, now is the time to test it against your real workloads.
Originally published at https://www.aimadetools.com
Top comments (0)