Skip to content
News Model release

Anthropic ships Claude Sonnet 5 as the new default model for Free and Pro plans at $2 per million input tokens

· by Pondero Newsdesk

The short version

Claude Sonnet 5 launched June 30 as the default model across Free and Pro tiers, beating Sonnet 4.6 on agentic coding benchmarks and approaching Opus 4.8 performance at Sonnet pricing.

Anthropic ships Claude Sonnet 5 as the new default model for Free and Pro plans at $2 per million input tokens

Anthropic shipped Claude Sonnet 5 on June 30 and immediately set it as the default model for Free and Pro users. The practical result: tens of millions of Claude users woke up to a model Anthropic describes as performing close to Opus 4.8 on agentic tasks, at roughly half the per-token cost of its predecessor.

What changed

Sonnet 5 replaces Sonnet 4.6 as the default across Free and Pro plans. It is also available to Max, Team, and Enterprise users. Developers can reach it via the API using the identifier claude-sonnet-5.

The announcement describes clear performance gains over Sonnet 4.6 on reasoning, tool use, coding, and knowledge work per Anthropic's announcement. The headline benchmark claim is that Sonnet 5, at higher effort levels, can match Opus 4.8 on certain agentic evaluations including BrowseComp (agentic search) and OSWorld-Verified (computer use). At medium effort levels it offers substantially better cost efficiency than Opus 4.8 while staying ahead of Sonnet 4.6 across the board.

Introductory API pricing is $2 per million input tokens and $10 per million output tokens, running through August 31, 2026. After that date, standard pricing applies at $3 per million input tokens and $15 per million output tokens. Anthropic noted the introductory rate is calibrated to make the transition from Sonnet 4.6 roughly cost-neutral, accounting for an updated tokenizer in Sonnet 5 that can produce 1.0 to 1.35 times more tokens for the same input depending on content type.

Opus 4.8 is priced at $5 per million input tokens and $25 per million output tokens for comparison.

Why it matters

For teams running agentic coding workflows, the core trade-off just shifted. Sonnet 4.6 sat clearly below Opus 4.8 on agentic tasks; Sonnet 5 narrows that gap enough that Anthropic now calls it "the recommended default for most production use cases."

In concrete terms: the model is available in Claude Code today. Early-access partners quoted in the announcement described Sonnet 5 completing multi-step software engineering tasks where Sonnet 4.6 stopped short, and self-checking its own output without being prompted. One engineer wrote that the model "wrote a reproducing test, implemented the fix, then stashed it to confirm the bug came back without the change. All in a single pass."

Tools like Cursor and Cline both expose Anthropic model selectors, so operators can switch to claude-sonnet-5 in their agent settings today. At the introductory price, it costs the same to run 2.5 Sonnet 5 calls as one Opus 4.8 call on inputs. That math tilts the default recommendation for most coding agent workloads toward Sonnet 5 until the September pricing kicks in.

Anthropic also noted safety improvements: Sonnet 5 showed lower rates of undesirable behaviors than Sonnet 4.6 on the company's automated behavioral audit, and the model resists prompt injection attacks more reliably in agentic settings. Cyber safeguards are enabled by default.

What to watch next

The August 31 introductory pricing cutoff is the most concrete near-term signal to track. At standard rates, the cost-per-call roughly doubles versus today. Operators who build cost assumptions around $2/$10 will need to revisit those numbers in September.

A Claude Code team post with a fuller agentic benchmark breakdown has not shipped as of publication. The full safety and capability evaluation data is in the Claude Sonnet 5 System Card.

Sources