Opus 5.5 monthly bill
$0
—
Claude Opus 5.5 just dropped
Opus 5.5 Cost & Savings Calculator
Anthropic cut input/output 20% and cache reads 60% vs Opus 5 — claiming ~40% lower cost on typical agent workloads. Plug in your tokens and see the bill, the savings, and how rivals compare.
Workload inputs
All prices are USD per million tokens (MTok). Defaults mirror a busy coding-agent month.
vs Opus 5 (same tokens)
$0
—
vs Opus 5 (+ efficiency)
$0
—
Annual savings (w/ efficiency)
$0
12 × monthly delta
Model cost board
Same token mix, list prices as of Sep 22, 2026. Cache columns use each vendor’s published cache-read / cache-write rates when available.
| Model | In / Out | Cache R / W | Est. month | vs Opus 5.5 |
|---|
Sources: Anthropic Opus 5.5 announcement (Sep 22, 2026), Anthropic platform pricing, OpenAI GPT-6 Astra docs, xAI Grok 4.7 announcement. Not affiliated with Anthropic, OpenAI, or xAI. Verify live pricing before budgeting.
Why the bill drops so hard
- Cache reads at $0.20/MTok — 60% cheaper than Opus 5’s $0.50. For agent loops that re-read huge contexts, this is the big lever.
- Input $4 / output $20 — 20% below Opus 5 ($5 / $25).
- Fewer tokens per task — Anthropic reports Opus 5.5 often finishes with fewer steps; the efficiency slider models that.
- Fast mode trade-off — $8 / $40 when you need up to ~2.5× speed.