· CostMyTokens · Analysis · 6 min read

Claude Opus 5.5 API Pricing (Nov 2026): $4/$20 per 1M Tokens

Claude Opus 5.5 API pricing: $4/$20 per 1M, cache read $0.20, Batch $2/$10, Fast $8/$40. ~$72/mo at 10M tokens. Official Anthropic rates vs Opus 5, Sonnet, GPT-6 Astra.

Quick answer: Claude Opus 5.5 costs $4 per million input tokens and $20 per million output tokens on the Claude API. Cached input reads cost $0.20 per million — 5% of the base input rate — and the Batch API halves everything to $2/$10. On our standard workload (10M tokens a month, 8:2 input/output split) that works out to $72 a month, and the full 1M-token context window bills at the standard rate with no long-context surcharge. Prices in US dollars, verified against Anthropic’s official pricing page on November 10, 2026. The live per-model calculator is on the Claude Opus 5.5 model page.

Disclosure: this article contains no affiliate links. Where CostMyTokens does use affiliate links elsewhere, they never affect order — every table is sorted by price only. Read our disclosure.

Claude Opus 5.5 prices at a glance

RatePrice per 1M tokens
Input$4
Output$20
Cache write (5-minute)$5
Cache write (1-hour)$8
Cache read$0.20
Batch input / output$2 / $10
Fast mode input / output$8 / $40

Two things stand out against the rest of the Claude lineup (full table on our Claude API pricing page). First, Opus 5.5 is priced below the previous Opus generation: Opus 5 sits at $5/$25 on the same official page, so 5.5 is a 20% cut on both input and output, not a premium launch price. Second, its cache read multiplier is 0.05x, half the standard 0.1x most models use — only the Fable tier reads cheaper (0.025x). For workloads that re-read the same long context every call, that multiplier matters more than the headline $4.

What it costs on real workloads

WorkloadPer callPer month
Chat assistant — 2,000 input / 500 output per call, 10,000 calls$0.018$180
Document processing — 50,000 input / 2,000 output per call, 1,000 calls$0.24$240
Agent loop with prompt caching — 10K fresh + 100K cached input / 2,000 output per call, 500 calls$0.10$50

The caching row is the one to study. The agent loop moves 110,000 input tokens per call, yet costs less per call than the document pipeline moving 52,000 — because 100K of those tokens bill at $0.20 instead of $4. At our standard workload, marking half the input as cache reads drops the month from $72 to $56.80; running the same workload through the Batch API instead drops it to $36. Run your own mix in the homepage calculator — it applies the cached and batch rates separately.

One metering note before you budget from someone else’s token counts: Opus 5.5 uses the newer Anthropic tokenizer, which the official page says produces roughly 30% more tokens for the same text than the previous tokenizer. A prompt that counted 100K tokens on an older Claude model may count ~130K here, so per-request costs rise even though the per-token rate fell versus Opus 5.

Where Opus 5.5 sits on price

ModelInputOutputStandard workload*
Claude Sonnet 5.5$2$10$36/mo
Claude Opus 5.5$4$20$72/mo
Claude Opus 5$5$25$90/mo
Claude Fable 5.1$10$50$180/mo
GPT-6 Astra$10$50$180/mo

* 10M tokens a month, 8:2 input/output split, no caching or batch discount.

Opus 5.5 is exactly 2x Sonnet 5.5 on both rates — the cleanest tier step in the current lineup — and exactly 40% of GPT-6 Astra’s rates, which is why it wins every workload in our head-to-head: see the full GPT-6 Astra vs Claude Opus 5.5 comparison and the wider Claude vs GPT-6 API pricing breakdown. The honest summary: you pay Opus 5.5 prices for the work Sonnet gets wrong, not as a default for all traffic — at 2x, routing easy calls down a tier halves that share of the bill.

The modifiers that change the bill

  • Prompt caching. Writes cost 1.25x base (5-minute TTL) or 2x base (1-hour TTL); reads cost 0.05x. The official guidance holds here: a 5-minute cache pays for itself after a single read, a 1-hour cache after two.
  • Batch API. 50% off input and output for asynchronous work that can wait — the cheapest way to run Opus 5.5 at volume.
  • Fast mode. $8/$40 — exactly double the standard rates — for latency-sensitive Opus 5.5 calls. It stacks with caching and data-residency multipliers and is not available with Batch.
  • US-only inference. Setting inference_geo: "us" applies a 1.1x multiplier to every token category. Global routing (the default) is standard price.
  • Long context. No surcharge: the full 1M window bills at standard rates, unlike Haiku 5.5, whose price steps up past 100K tokens of prompt.

FAQ

How much does Claude Opus 5.5 cost per 1M tokens? $4 per 1M input tokens and $20 per 1M output tokens. Cache reads cost $0.20 per 1M, cache writes $5 (5-minute) or $8 (1-hour), and Batch API rates are $2 input / $10 output. Verified against Anthropic’s official pricing page on November 10, 2026.

What does Claude Opus 5.5 cost per month? It depends on volume. On a standard workload of 10M tokens a month (8M input, 2M output) it costs $72 at standard rates, $56.80 if half the input is cache reads, or $36 on the Batch API. A chat assistant making 10,000 calls a month costs about $180.

Is Claude Opus 5.5 cheaper than Claude Opus 5? Yes. Opus 5.5 is $4/$20 per 1M tokens versus $5/$25 for Opus 5 — 20% less on both input and output, per Anthropic’s official pricing page.

How does Opus 5.5 compare to Sonnet 5.5 on price? Opus 5.5 costs exactly twice Sonnet 5.5 on both rates: $4/$20 versus $2/$10 per 1M tokens, with the same 0.05x cache-read multiplier structure (cache reads $0.20 vs $0.10).

Does the 1M context window cost extra on Opus 5.5? No. Anthropic bills the full 1M-token context window at standard rates — a 900K-token request pays the same per-token price as a 9K-token one. The only geographic modifier is US-only inference (inference_geo: "us"), which adds a 1.1x multiplier.

How much does prompt caching save on Opus 5.5? Cache reads bill at 0.05x the base input rate — $0.20 instead of $4 per 1M tokens, a 95% saving on the cached share. Cache writes cost more upfront (1.25x for 5-minute, 2x for 1-hour), so caching pays off when the same context is read at least once (5-minute) or twice (1-hour).

Price Opus 5.5 on the Claude Opus 5.5 model page, browse the full Anthropic lineup on Claude API pricing, and run your own volumes on the cost calculator. Head-to-head: GPT-6 Astra vs Claude Opus 5.5 and Claude vs GPT-6 API pricing.

Prices verified against Anthropic’s official pricing page on November 10, 2026, and cross-checked with the Claude Opus 5.5 model page tracked on this site. Anthropic states prices in USD per million tokens.

Share:
Back to Blog

Related Posts

View All Posts »