Claude Sonnet 4.6 Pricing, Plus Sonnet 5.5 and 5 Costs
Claude Sonnet 4.6 pricing next to Sonnet 5.5, Sonnet 5 and 4.5: API cost per million tokens, caching, batch, a worked example and which plans include Sonnet.
Josh Pocock7 min read
AI Coding
Sonnet pricing
Stride AI Academy
Claude Sonnet 4.6 pricing is $3 per million input tokens and $15 per million output tokens on the Claude API. The newer Sonnet 5.5 and Sonnet 5 are cheaper per token at $2 input and $10 output. Sonnet 4.5 costs the same as 4.6 but is deprecated and retires on November 30, 2026.
This guide has the full Sonnet price table, caching and batch rates, what happened to older Sonnet models, a worked example that accounts for the newer tokenizer, and where Sonnet sits in each Claude plan. Everything here was checked against Anthropic's pricing docs, models overview, Sonnet 5.5 overview, model deprecations, rate limits and Claude's pricing page on November 22, 2026. Prices are in US dollars per million tokens (MTok).
Claude Sonnet API pricing by model
From the pricing docs on November 22, 2026:
| Model | Input | Output | 5-min cache write | 1-hour cache write | Cache hit |
|---|---|---|---|---|---|
| Sonnet 5.5 | $2 | $10 | $2.50 | $4 | $0.10 |
| Sonnet 5 | $2 | $10 | $2.50 | $4 | $0.20 |
| Sonnet 4.6 | $3 | $15 | $3.75 | $6 | $0.30 |
| Sonnet 4.5 (deprecated) | $3 | $15 | $3.75 | $6 | $0.30 |
| Sonnet 4 (retired) | $3 | $15 | $3.75 | $6 | $0.30 |
Sonnet 5.5's cache hits cost 0.05x its input price. The others use the standard 0.1x.
Sonnet 5 pricing is now permanent
Sonnet 5 launched at $2 and $10 as introductory pricing through August 31, 2026. The pricing docs say that rate is now the standard price, and the planned rise to $3 and $15 on September 1, 2026 won't happen.
Claude Sonnet 4.5 pricing and retirement
Sonnet 4.5 costs $3 input and $15 output, the same as 4.6. The deprecations page says Anthropic notified developers on September 30, 2026 that it retires on the Claude API on November 30, 2026, with claude-sonnet-5-5 as the recommended replacement. That includes the dateless claude-sonnet-4-5 ID. Once it's retired, requests to it fail.
Claude 3.5 Sonnet pricing
Claude 3.5 Sonnet is no longer on the pricing page. The deprecations page lists both 3.5 Sonnet models as retired on October 28, 2025, and Claude Sonnet 3.7 as retired on February 19, 2026. The recommended replacement for each is Sonnet 5.5. Sonnet 4 retired on June 15, 2026.
Sonnet caching and batch pricing
Caching. A 5-minute cache write costs 1.25x the input price and a 1-hour write costs 2x. The pricing docs say a 5-minute cache pays for itself after one read and a 1-hour cache after two.
Batch API. 50% off input and output:
| Model | Batch input | Batch output |
|---|---|---|
| Sonnet 5.5 | $1 | $5 |
| Sonnet 5 | $1 | $5 |
| Sonnet 4.6 | $1.50 | $7.50 |
| Sonnet 4.5 | $1.50 | $7.50 |
Caching and batch discounts stack. Two more multipliers to know: US-only inference with inference_geo: "us" costs 1.1x on Claude 4.6 and later models, and Sonnet 4.6, 5 and 5.5 include the full 1M-token context window at standard pricing. For the same rules across every model, see the Claude API pricing guide.
Sonnet 4.6 vs Sonnet 5.5: which costs less?
Per token, Sonnet 5.5 is a third cheaper than Sonnet 4.6. But the pricing docs say Claude 4.7 and later models use a newer tokenizer that produces about 30% more tokens for the same text. Sonnet 4.6 uses the older one, and Sonnet 5 and 5.5 use the newer one. So the same prompt makes more tokens on Sonnet 5.5.
Here's a rough worked example. Say a job uses 1 million input tokens and 100,000 output tokens on Sonnet 4.6:
| Line | Sonnet 4.6 | Cost |
|---|---|---|
| Input | 1 MTok × $3 | $3.00 |
| Output | 0.1 MTok × $15 | $1.50 |
| Total | $4.50 |
If the same text comes to about 30% more tokens on Sonnet 5.5, that's roughly 1.3 million input and 130,000 output:
| Line | Sonnet 5.5 | Cost |
|---|---|---|
| Input | 1.3 MTok × $2 | $2.60 |
| Output | 0.13 MTok × $10 | $1.30 |
| Total | $3.90 |
Sonnet 5.5 still comes out about 13% cheaper here. The docs say the exact increase depends on your content, so count your own prompts with the token counting endpoint and each model's ID before you decide. Output length also differs between models, which no table can predict.
Caching widens the gap. A cache hit costs $0.10 on Sonnet 5.5 against $0.30 on Sonnet 4.6, so long agent sessions with a big cached prefix favour 5.5 even more.
Sonnet specs that affect your bill
From the models overview and the Sonnet 5.5 page:
| Sonnet 5.5 | Sonnet 5 | Sonnet 4.6 | Sonnet 4.5 | |
|---|---|---|---|---|
| Context window | 1M | 1M | 1M | 200K |
| Max output | 128K | 128K | 128K | 64K |
| Start tier RPM | 4,000 | 1,000 | 1,000 (shared with 4.5) | 1,000 (shared with 4.6) |
The rate limits docs say Sonnet 4.6 and 4.5 share one combined limit, while Sonnet 5.5 and Sonnet 5 each have their own. Our guide to Claude API rate limits has every tier.
Sonnet 5.5's default effort is high. It also has two behaviour changes to check before you migrate. Its thinking can't be fully switched off: the lowest setting is between_tools, which turns off up-front thinking. And setting temperature, top_p or top_k returns a 400 error. Effort is the main cost lever. The choosing-a-model docs say tuning effort is often a better lever than switching models.
Where Sonnet sits in each Claude plan
On a subscription, there's no per-token charge. Claude's pricing page lists Sonnet on every plan on November 22, 2026, including Free:
| Plan | Price | Sonnet |
|---|---|---|
| Free | $0 | Yes, with limited usage |
| Pro | $20/month, or $17/month billed yearly | Yes |
| Max | $100/month | Yes, 5x Pro usage |
| Ultra | $200/month | Yes, 10x Pro weekly usage |
| Team | From $20/seat/month billed yearly | Yes |
| Enterprise | $20/seat/month billed yearly plus usage at API rates | Yes |
Opus is the model that Free doesn't get. See Claude pricing for the full plan breakdown.
In Claude Code, the model configuration docs say the sonnet alias points to Sonnet 5.5 on the Anthropic API, and opusplan uses Opus in plan mode, then Sonnet for execution.
Sonnet vs the rest of the lineup
| Model | Input | Output |
|---|---|---|
| Fable 5.2 | $8 | $40 |
| Opus 5.5 | $4 | $20 |
| Sonnet 5.5 | $2 | $10 |
| Haiku 5.5 | From $0.10 | From $0.50 |
The models overview describes Sonnet 5.5 as "the best combination of speed and intelligence." Opus 5.5 costs twice as much and is the docs' suggested starting point for most workloads. Haiku 5.5 is far cheaper for high-volume work. Compare them in Claude Opus pricing and Claude Haiku vs Sonnet.
Learn to build with Claude Code
The Stride AI Coding course teaches Claude Code and Codex for building client tools, written for non-programmers, with weekly live calls to review your builds. Many of the free guides cover Claude Code setups, and the Claude Code community is where members compare theirs.
FAQ
What is Claude Sonnet 4.6 pricing?
Claude Sonnet 4.6 costs $3 per million input tokens and $15 per million output tokens on the Claude API. Cache hits cost $0.30 and the Batch API halves the rates to $1.50 and $7.50.
Is Sonnet 5.5 cheaper than Sonnet 4.6?
Per token, yes: $2 and $10 against $3 and $15. Sonnet 5.5 uses a newer tokenizer that makes about 30% more tokens for the same text, so the real saving is smaller than the price gap. Count your own prompts to compare.
What is Claude Sonnet 4.5 pricing?
Sonnet 4.5 costs $3 input and $15 output per million tokens, the same as 4.6. It's deprecated and retires on the Claude API on November 30, 2026, with Sonnet 5.5 as the recommended replacement.
How much does Claude 3.5 Sonnet cost?
You can't buy it any more. Both Claude 3.5 Sonnet models retired on October 28, 2025, and they're no longer on the pricing page. Anthropic recommends Sonnet 5.5 instead.
Will Sonnet 5 pricing go up?
No. The pricing docs say the $2 and $10 introductory price for Sonnet 5 is now the standard price, and the planned increase to $3 and $15 on September 1, 2026 won't happen.
Is Sonnet free on the Claude app?
Yes, with limited usage. Claude's pricing page lists Sonnet on the Free plan as well as on Pro, Max, Ultra, Team and Enterprise.