Anthropic launches Claude Sonnet 5 at a steep discount to Opus as it races toward IPO
Anthropic today released Claude Sonnet 5, a mid-tier model the company says delivers near-flagship performance at mid-tier prices, timed to build broad developer adoption just as the San Francisco lab barrels toward an IPO that will test whether private AI valuations survive public scrutiny. Sonnet 5 is now the default model for Free and Pro users and is available to Max, Team, and Enterprise customers. Introductory API pricing is $2 per million input tokens and $10 per million output tokens through August 31, rising afterward to $3 and $15 respectively — still far below the $5 input / $25 output pricing of Opus 4.8.
On benchmarks, Sonnet 5 posts major gains over predecessor Sonnet 4.6 across every disclosed evaluation. On SWE-bench Pro, an agentic coding benchmark, it scores 63.2% versus Sonnet 4.6's 58.1%, and Anthropic says it narrowed the gap with Opus across five major evaluations and surpassed it on one. The model is pitched squarely at cost-conscious enterprise developers who want powerful agentic capabilities — planning, tool use, and autonomous multi-step execution — without paying flagship prices.
The strategic logic is unmistakable: Anthropic is democratizing capabilities that until recently only its most expensive models could deliver, the kind of broad-based developer traction that looks attractive in an S-1 filing. The launch also coincides with the company's separate push into vertical products like Claude Science and its Claude Code tooling, where Claude Code lead Boris Cherny previewed background-running subagents.
Skeptics flagged pricing subtleties: independent commentator Simon Willison noted Sonnet 5's new tokenizer makes it roughly 1.4x more expensive for English and 1.33x for Spanish, though about the same for Simplified Mandarin — meaning the headline per-token discount understates real-world English costs. Watch whether the introductory pricing sticks after August 31 and whether Sonnet 5 cannibalizes Opus revenue as it closes the benchmark gap.