Last verified
ARCHIVE PRICERETIRED MODEL200KTEXT + VISIONPROMPT CACHING

Claude 3.5 Sonnet API Pricing

Original Claude 3.5 Sonnet is retired; this archive preserves the $3/$15 launch price. $3/M input, $15/M output, and $0.3/M cache-hit input. This was the June 2024 Sonnet release that made the $3 input and $15 output tier a major production baseline.

Archived input - per 1M tokens
$3.00/M
Source Anthropic retired
Archived output - per 1M tokens
$15.00/M
Use for invoice checks retired
Cached input - 90% off
$0.30/M
Cache reads -90%
Effective - agentic blend
$1.92/M
92/8 split - 82% cache
§ 01 / TERMINAL

Run the numbers.

Archive calculator pre-loaded with Claude 3.5 Sonnet rates. Tweak spend, output mix, or cache hit rate to compare historical bills with current replacements.

$ /mo
Workload split
Prompt cache hit rate
Tokens you can process
—
Words equivalent (English)
—
Effective rate
—
§ 02 / SCENARIOS

Real-world presets.

CODING AGENT RETIRED

Repo-wide bug fix

$0.169/task
81k in - 7k out~591 units/$100
LONG DOC ANALYSIS RETIRED

Reading 100-page contracts

$0.258/doc
175k in - 8k out~387 units/$100
RAG SUPPORT RETIRED

Support agent ticket triage

$0.018/ticket
4k in - 1k out~5,555 units/$100
ASSISTANT TURN RETIRED

Research planning turn

$0.039/turn
12k in - 2k out~2,564 units/$100
§ 03 / TAPE

Price history.

Claude 3.5 Sonnet is retired; archived at $3/$15 per M.

Input · $3/M
Output · $15/M
Cached · $0.30/M
JUN 20 Launch at $3/M input and $15/M outputMAY 18 Retired; replacement is Claude Sonnet 4.6
§ 04 / TOKENIZER

Paste text. See tokens. See cost.

Estimate · anthropic-bpe-estimate · ≈3.5 chars/token Auto-counts as you type

This is a chars-per-token approximation, not a real tokenizer. Actual tokens vary by language, code density, and tool-call overhead — counts are typically ±10–20% off for English prose, more for code or non-Latin scripts. For exact billing, use the vendor's official tokenizer.

Characters —
Words —
Tokens (estimated) —
Cost as input · uncached —
Cost as output · uncached —
Cost as cached input —
§ 05 / SHELF

Up against the shelf.

All models →
Model Input /M Output /M Effective blended Context Best for
Claude Sonnet 3.5 (June 2024) $3.00 cache $0.30 $15.00 $1.92 current archive 200K Production Claude agents
Claude Sonnet 4.6 $3.00 cache $0.30 $15.00 $1.92 same blend 1M Production Claude agents
Claude Sonnet 4.5 $3.00 cache $0.30 $15.00 $1.92 same blend 200K Production Claude agents
Claude Haiku 4.5 $1.00 cache $0.10 $5.00 $0.64 cheaper 200K Low-latency Claude workloads
Claude Opus 4.7 $5.00 cache $0.50 $25.00 $3.21 pricier 1M Frontier reasoning and hard code
GPT-5.4 $2.50 cache $0.25 $15.00 $1.80 same blend 1.05M OpenAI production apps
Gemini 2.5 Pro $1.25 cache $0.13 $10.00 $1.10 cheaper 2M Gemini long-context work
DeepSeek V4 Pro $0.43 cache $0.00 $0.87 $0.14 cheaper 1M Low-cost reasoning and coding

Frequently asked.

Short answers for teams checking old Claude 3.5 Sonnet costs or migrating archive workloads.

Q · 01 Is Claude 3.5 Sonnet still available? +
Claude 3.5 Sonnet is retired on the first-party Claude API. Keep this page for archive pricing, old invoices, and migration planning.
Q · 02 What replaced Claude 3.5 Sonnet? +
Anthropic lists claude-sonnet-4-6 as the recommended replacement in its model deprecations page.
Q · 03 How much did Claude 3.5 Sonnet cost? +
Claude 3.5 Sonnet is archived at $3/M input and $15/M output.
Q · 04 Does this archive include prompt caching? +
Cache-hit input is archived at $0.3/M, matching Anthropic's 10% cache-read multiplier on this price tier.
Q · 05 Can I use this for new production routing? +
Use the archive for cost history, not fresh routing decisions. Compare the current Claude rows in the shelf before starting new workloads.
Q · 06 Why keep a page for a retired model? +
Retired-model pages help teams reconcile historical spend, understand migration economics, and answer old search demand without pretending the model is current.
Y
Reviewed by Yaroslav Vikhariev Founder - AI//COST - Last verified 2026-05-18 - Anthropic archive pricing
Methodology Report a correction More by Y.V.