The awkward fact about Claude Opus 4.6 is that its successor is cheaper. Opus 4.6 lists at $5 per million input tokens and $25 output; Claude Opus 5.5, which Anthropic now recommends for new work, is $4 and $20.
So if you're still on 4.6, the useful question is what keeps you there. Released February 5, 2026 as claude-opus-4-6, it's marked Active (legacy) in Anthropic's docs, with retirement no sooner than February 5, 2027.
Claude Opus 4.6 Pricing
From Anthropic's pricing page on October 12, 2026, USD per million tokens:
| Item | Price |
|---|---|
| Input / output | $5 / $25 |
| Cache write, 5 min / 1 hour | $6.25 / $10 |
| Cache read | $0.50 |
| Batch API | Half price: $2.50 / $12.50 |
US-only inference (inference_geo: "us") adds 1.1× on Claude 4.6 and later.
Claude Opus 4.6 1M Context
The full 1M-token window is generally available on the Claude Platform, Amazon Bedrock, Google Cloud Vertex AI and Microsoft Foundry, billed at the same $5/$25 all the way through. There's no long-context surcharge, and Claude Code Max, Team and Enterprise sessions on Opus 4.6 open the whole window by default.
Max output is 128K tokens, or 300K through the Batch API with a beta header. Thinking is adaptive (extended thinking is deprecated) with high as the default effort. Knowledge cutoff is May 2025, training data August 2025. The model ID is claude-opus-4-6 everywhere except Bedrock's InvokeModel, which uses anthropic.claude-opus-4-6-v1.
Arena Results
In Arena's October 8, 2026 text snapshot (style control), claude-opus-4-6-high ranked third overall at 1504.4 (1501.0–1507.7) on 79,415 votes, behind only Gemini 4 Argon high (1525.4) and Opus 5.5 high (1507.1). The default entry, claude-opus-4-6, ranked seventh at 1497.7 (1494.4–1501.0) on 84,095 votes.
| Category | High | Default |
|---|---|---|
| Coding | #4, 1549.7 (20,617 votes) | #6, 1548.1 (23,351) |
| Instruction following | #3, 1513.3 (25,594) | #6, 1499.5 (28,177) |
| Multi-turn | #2, 1517.8 (13,495) | #6, 1512.1 (14,773) |
| Creative writing | #4, 1500.3 (14,532) | #12, 1478.8 (14,689) |
| Chinese | #5, 1550.4 (5,496) | #8, 1544.7 (5,735) |
| Math | #6, 1518.3 (4,085) | #9, 1506.8 (4,570) |
Effort matters most for creative writing. Ranks two through six overlap, so the exact order there isn't settled.
→ Track it on AI Rank: Claude Opus 4.6 high — Arena rating, rank and 90-day history
Traffic tells the other half. On OpenRouter, anthropic/claude-4.6-opus-20260205 (shown with "(batch)") ranked #79 for October 4–10 at about 91.7B tokens (+10%), #70 on October 10 at about 14.5B (−2%), and #82 for the month ending October 10 at about 404B (+378%). Newer Opus and Fable models sit far above it on the OpenRouter usage ranking on AI Rank.
If claude-opus-4-6 is hard-coded in a pipeline, there's no rush: it scores well and has more than a year before retirement. For anything new, Opus 5.5 costs less and ranks a little higher. Check the current Arena ratings before you commit.
Compare before you choose: pricing changes and rankings move daily — check the current model rankings on AI Rank before you commit.
Or 📱 get the AI Rank app to follow them on your phone.
Prepared with automated editorial assistance and checked against official sources and dated AI Rank data on October 12, 2026. We did not run our own benchmarks for this article.