Claude 3.5 Sonnet: Pricing, Context, and Benchmarks
Claude 3.5 Sonnet is Anthropic's Claude 3 model released on 2024-10-22. It runs on a 200K-token context window, priced at $3.00/M input and $15.00/M output. Previous-gen flagship. Still widely deployed; deprecation timeline announced for late 2026.
Specifications
| Provider | Anthropic |
|---|---|
| Family | Claude 3 |
| Released | 2024-10-22 |
| Status | active |
| Context window | 200K tokens |
| Max output | 8K tokens |
| Modalities | text image |
| Capabilities | function calling json mode vision streaming tool use |
| Tokenizer | claude-3 |
| API endpoint | https://api.anthropic.com/v1/messages |
| Documentation | Official docs ↗ |
| Last updated | 2026-05-29 |
Cost calculator
For 1,000 tokens of input and 1,000 tokens of output per call, this model costs $0.0180 per call. For typical usage:
| Workload | Tokens / call (in/out) | $/call | $/1K calls | $/1M calls |
|---|---|---|---|---|
| Short Q&A | 200 / 100 | $0.002100 | $2.10 | $2,100 |
| Standard chat | 1K / 500 | $0.010500 | $10.50 | $10,500 |
| RAG with retrieval | 4K / 500 | $0.019500 | $19.50 | $19,500 |
| Long doc summary | 20K / 1K | $0.075000 | $75.00 | $75,000 |
| Long context (100K input) | 100K / 2K | $0.322500 | $322.50 | $322,500 |
For more precise estimates including caching discounts, use the LLM API Cost Calculator.
Other Anthropic models
Common comparisons
- Claude 3.5 Sonnet vs Grok 3 Side-by-side specs, pricing, and benchmarks
- Claude 3.5 Sonnet vs GPT-4o Side-by-side specs, pricing, and benchmarks
- Claude 3.5 Sonnet vs Llama 3.1 405B Side-by-side specs, pricing, and benchmarks
- Claude 3.5 Sonnet vs Command R+ Side-by-side specs, pricing, and benchmarks
API setup
curl https://api.anthropic.com/v1/messages \
-H "Authorization: Bearer $API_KEY" \
-H "Content-Type: application/json" \
-d '{"model": "claude-3-5-sonnet", "messages": [{"role": "user", "content": "Hello"}]}'See the official documentation for the full request/response schema. For pre-flight token counting before paying, use the appropriate token counter.