Claude Sonnet 5.5 vs Opus 5.5: pricing, specs and how to choose
Both ship with a 1M context and 128K maximum output: Sonnet 5.5 costs $2 input / $10 output, Opus 5.5 $4 / $20, and cache reads are $0.10 and $0.20 (Sonnet 5.5's were cut from $0.20 on 2026-10-07). In the official model overview of 2026-10-20 both are still current models, with retirement not sooner than 2027-09-28 and not sooner than 2027-09-22 respectively. The official wording states each model's role; neither is ranked above the other.
Updated 2026-10-20
Switch models on the same endpoint and measure the cost per token yourself; check /models for callable models.
- Billed per token — live rates on /models
- Pay by card (Visa / Mastercard), Apple Pay, Google Pay or crypto
- Self-service: top up and activate instantly
Four key points
Same context and output
Both are listed at 1M context with 128K maximum output in the official model docs, so the choice never buys you a longer window.
Sonnet 5.5 rates
$2 per million input and $10 per million output, the same as Sonnet 5 per Anthropic; cache reads at $0.10 since 2026-10-07 ($0.20 at launch; Sonnet 5 stays at $0.20).
Opus 5.5 rates
$4 per million input and $20 per million output, with $0.20 cache reads; converted from the official table, Sonnet 5.5 sits at exactly half on all three lines (on cache reads since 2026-10-07).
Switching changes one field
As of 2026-09-29 both are callable on QCode: same endpoint, same key, only the model field changes, so a side-by-side run is a one-line edit.
What actually differs
The limits match: 1M context and 128K maximum output for both. Unit price and stated role differ: $2/$10 for Sonnet 5.5 against $4/$20 for Opus 5.5, so converted from the official table Sonnet 5.5 is exactly half on both sides, and since 2026-10-07 on cache reads too ($0.10 against $0.20 per million). Anthropic describes Opus 5.5 as built for complex work requiring careful judgment and Sonnet 5.5 as strongest at well-scoped everyday tasks. The official model overview of 2026-10-20 adds one more reference point: if you are unsure, start with Opus 5.5, and use Fable 5.1 for demanding reasoning and long-horizon agentic work.
Six days apart at launch
Anthropic released Opus 5.5 on 2026-09-22 and Sonnet 5.5 on 2026-09-28, with the launch post stating at release that Sonnet 5.5 is priced the same as Sonnet 5 ($2 input, $10 output, $0.20 cache reads; the page now reads $0.10); on 2026-10-07 the official release notes cut Sonnet 5.5 cache reads to $0.10. Both carry the same 1M context and 128K maximum output, so the difference sits in unit price and stated role rather than context length. Two launch-page sentences were added on the 2026-10-20 check: Anthropic writes that Opus 5.5 performs at the level of Fable 5.1 on most work and costs 40% less to run than Opus 5, and that Sonnet 5.5 runs 30% faster than Sonnet 5 and costs up to 30% less for most work.
Public timeline
On 2026-09-22 Anthropic released Claude Opus 5.5; the price table lists $4 input / $20 output with cache reads at $0.20, which the launch post describes as 20% below Opus 5 on tokens and 60% below on cache reads.
On 2026-09-28 Anthropic released Claude Sonnet 5.5, stating it is priced the same as Sonnet 5: $2 per million input tokens, $10 per million output, $0.20 per million for cache reads (cut to $0.10 on 2026-10-07).
Check date: as of 2026-09-29 both models are callable on QCode through the same Anthropic-compatible endpoint (model ids claude-sonnet-5-5 and claude-opus-5-5), and both show a non-zero call count in the trailing thirty-day record taken that day.
Confirmed vs unverified
Confirmed (official docs)
Re-checked on 2026-10-20 against Anthropic's model docs and price table, confirmed: Sonnet 5.5 - 1M context, 128K maximum output, $2 input / $10 output, cache writes at $2.50 (5-minute) and $4 (1-hour), cache reads $0.20 at the time (cut to $0.10 on 2026-10-07, per the official release notes); Opus 5.5 - 1M context, 128K maximum output, $4 input / $20 output, cache writes $5 and $8, cache reads $0.20. The default-effort row lists high for Sonnet 5.5 and medium for Opus 5.5. The same overview of 2026-10-20 also lists a Jun 2026 knowledge cutoff for both, and retirement not sooner than 2027-09-28 (Sonnet 5.5) and not sooner than 2027-09-22 (Opus 5.5).
Unverified · not carried here
As of 2026-09-29, two things are left out of this page: any verdict on which model is stronger — there is no official head-to-head, and the wording only names the work each one is aimed at — and every benchmark figure or rumour about the pair, including tables in the launch posts, since those move between releases and stay unconfirmed by any independent run. What stays are the unit prices from one price table and the length limits from one model document.
Line-by-line
Sonnet 5.5 and Opus 5.5 (official rates)
Input $2 against $4, output $10 against $20, 5-minute cache writes $2.50 against $5, 1-hour cache writes $4 against $8, cache reads $0.10 against $0.20 (all per million tokens; Sonnet 5.5's cache reads were also $0.20 until the official 2026-10-07 cut). Context is 1M and maximum output 128K on both sides. Converted from the official table: exactly half on input, output and cache writes, and since 2026-10-07 on cache reads too.
Stated role vs unit-price ratio
In Anthropic's words: Opus 5.5 is built for complex work requiring careful judgment, while Sonnet 5.5 is strongest at well-scoped everyday tasks, fixing bugs, and creating polished documents, slides, and spreadsheets. This page only places that wording next to the price table; it cites no benchmarks and draws no ranking.
How to choose, where to start
Start from scope: well-scoped, high-volume, cost-sensitive work points at Sonnet 5.5; long chains of judgement, cross-file refactors or expensive conclusions point at Opus 5.5. Anthropic's own advice — if you are unsure which model to use, start with Opus 5.5 — still stands. Both models sit on the same endpoint behind the same key, so a common pattern is to run Sonnet 5.5 first and switch the model field to Opus 5.5 where judgement gets tight.
Using them on QCode
As of 2026-10-20 both claude-sonnet-5-5 and claude-opus-5-5 are callable on QCode over the same Anthropic-compatible endpoint: copy a key in the console, keep the base URL, and put either model id in the model field. Only the model field changes; the rest of the request stays as it is. Per-model unit prices on QCode are on /models.
Questions
Are the specs identical?
On length they match: the official model docs give both 1M context and 128K maximum output. The difference is in unit price and stated role, not in the window.
How big is the price gap?
Per Anthropic's official table, per million tokens Sonnet 5.5 costs $2 input / $10 output and Opus 5.5 $4 / $20; cache reads are $0.10 and $0.20 (Sonnet 5.5's were also $0.20 until the official 2026-10-07 cut) — an exact two-to-one on input, output and cache reads (that ratio is our arithmetic from the table, not a vendor statement).
How does caching price out?
From the official table, per million tokens: Sonnet 5.5 has $2.50 for 5-minute cache writes, $4 for 1-hour writes and $0.10 for reads (since 2026-10-07; $0.20 before); Opus 5.5 has $5, $8 and $0.20. Opus 5.5 now costs twice as much on both writes and reads.
Do I need a new key or endpoint?
No. As of 2026-09-29 both models sit behind the same Anthropic-compatible endpoint on QCode: base URL, authentication and every other parameter stay put, and you swap claude-sonnet-5-5 for claude-opus-5-5 in the model field (or the other way round).
Are the default efforts the same?
No. The Default effort row of the official model table reads, in column order: high for Sonnet 5.5 and medium for Opus 5.5 (the Haiku 4.5 cell on that row says Not supported). The launch post adds two separate statements: in Claude Code and the Claude apps the default effort is Medium, while the Claude Platform defaults to High.
Can you say which one is stronger?
No. This page compares only the official price table, the model docs and the wording Anthropic uses for each model's role. Benchmarks and rankings are deliberately left out — they move between generations, while unit prices and context lengths are stated plainly. To decide, run your own task twice with the same code and only the model field changed.
Sources
Anthropic's official price table (the Sonnet 5.5 and Opus 5.5 rows), the official model document (context, maximum output, default effort, knowledge cutoff, retirement), the Claude Opus 5.5 and Claude Sonnet 5.5 launch pages (pricing and role wording), the official release notes of 2026-10-07 (the Sonnet 5.5 cache-read cut), plus the official pages captured on 2026-10-20.
Run both before deciding
Same key, same endpoint, only the model field moves: both Sonnet 5.5 and Opus 5.5 carry a 1M context and a 128K maximum output (as of 2026-10-20).
Related
Claude Sonnet 5.5: the full guide
The cheaper of the pair: $2/$10, 1M context, and the vendor's own speed and cost wording.
Claude Fable 5.2 vs Claude Opus 5.5 (2026-11)
Official comparison of Fable 5.2 and Opus 5.5: release dates, positioning, context, max output, official prices, thinking behavior and API differences, plus a selection guide built on the official statements and Claude Code switch commands.
Claude Opus 4.8 fast mode error and migration (removal 2026-11-30)
Checked 2026-11-05: deprecation and removal dates for Opus 4.8 fast mode, what speed: fast does afterwards, two fixes and the move to Opus 5.5.
Crypto payment guide
Nine coins, 56 networks, zero platform fee for AI coding API credits.
QCode is independent of Anthropic. Every unit price and length limit here comes from the Anthropic price table and model document captured on 2026-10-20; later official changes are written in one by one, for instance the official release notes of 2026-10-07, which cut Sonnet 5.5 cache reads to $0.10, while other repricing is not written back automatically. The ratios and worked examples are our arithmetic from that table. On the QCode side the page asserts only one thing: both models were callable as of 2026-10-20, and it does not restate Anthropic's list prices as ours.