Anthropic’s Claude Opus 4.7 ups coding chops — but an updated tokenizer can bloat token counts

A sharper coder, with eyes for design

It has been reported that Anthropic has released Claude Opus 4.7, the latest iteration of its Opus line aimed squarely at advanced software engineering tasks. The company says Opus 4.7 is noticeably better on the hardest coding problems — the stuff that used to need hands-on human supervision — and that it handles long, multi-step jobs with more rigor and self-checks. It has also been reported that the model brings improved vision and a more tasteful creative output for interfaces, slides and docs, and that Anthropic’s benchmarks put Opus 4.7 ahead of Opus 4.6, GPT-5.4 and Gemini 3.1 Pro, though still behind the invite-only Claude Mythos Preview.

Tokenizer tradeoff: better processing, more tokens

It has been reported that Opus 4.7 uses an updated tokenizer that “improves how the model processes text,” but “the tradeoff is that the same input can map to more tokens” — roughly 1.0–1.35× depending on content. Anthropic also notes the model "thinks more" at higher effort levels, especially in agentic settings, which increases output token volume. So yes: better thinking, but potentially higher token bills. Irony? Maybe. Box’s Head of AI, Yashodha Bhavnani, reportedly found Opus 4.7 delivered efficiency wins — fewer model and tool calls, faster responses and lower AI Unit usage — a counterpoint to the token-growth worry that enterprises will be parsing carefully.

Cadence and developer tools

Anthropic has been shipping Opus updates on a roughly two-month cadence, and it has been reported that the company published migration guidance and a new system card alongside the launch. Meanwhile, Claude Code in the Claude Mac app continues to evolve: auto mode has been expanded to Max plan subscribers, and recent tweaks give the assistant more Mac-level autonomy while aiming to keep safety in check. Developers and product teams now face a familiar trade-off — improved capability and convenience, against potentially higher operating costs and the classic “did it break my quota?” anxiety. Worth it? For teams wrestling with gnarly code, the relief might be palpable. For everyone else: read the migration notes and keep an eye on your token meter.

Sources: 9to5mac.com