Curiosity · AI Model
Claude 3.5 Sonnet
Claude 3.5 Sonnet is the June 2024 model that changed Anthropic's trajectory. It beat Opus 3 at a fraction of the cost, set a new SWE-bench high, and shipped the first public computer-use preview — turning Claude from 'the safety lab's chatbot' into the default model for developers building coding agents.
Model specs
- Vendor
- Anthropic
- Family
- Claude 3.5
- Released
- 2024-06
- Context window
- 200,000 tokens
- Modalities
- text, vision, code
- Input price
- $3/M tok
- Output price
- $15/M tok
- Pricing as of
- 2026-04-20
Strengths
- Set the 'good at coding' perception that still defines Claude today
- Huge quality gain over Opus 3 at the Sonnet price point
- First public computer-use demo, paving the way for agent UX
- Excellent tool-use stability for 2024-era models
Limitations
- No extended thinking — 3.7 Sonnet introduced that later
- 200K context; smaller than Claude 4 family
- Vision OCR below Gemini 2.5 Pro quality
- Largely superseded by Claude 4 models in 2025-26
Use cases
- Coding assistants — Cursor, Zed, Aider defaulted to it in 2024
- Computer-use preview workflows (Claude 3.5 Sonnet v2)
- Multi-turn chat agents with tool use
- RAG pipelines across long documents
Benchmarks
| Benchmark | Score | As of |
|---|---|---|
| SWE-bench Verified | ≈49% | 2024-10 |
| HumanEval | ≈93.7% | 2024-06 |
| MMLU | ≈88.7% | 2024-06 |
Frequently asked questions
What is Claude 3.5 Sonnet?
Claude 3.5 Sonnet is Anthropic's June 2024 model and the first to beat Claude Opus 3 at a cheaper tier. It defined Claude's reputation as a strong coding model, set new SWE-bench records at launch, and shipped the first public computer-use preview.
What was the 'Claude 3.5 Sonnet v2' update?
In October 2024 Anthropic quietly updated Claude 3.5 Sonnet with better coding, tool use, and the first public computer-use preview — informally called 'Claude 3.5 Sonnet v2'. It raised SWE-bench scores into the high-40s without a version bump.
Is Claude 3.5 Sonnet still worth using?
For new projects, use Claude Sonnet 4.6 or Sonnet 4.5 instead — they are faster, cheaper per unit of quality, and support 1M context and extended thinking. Claude 3.5 Sonnet is still offered on the API for pipelines tuned to its behaviour.
How much did Claude 3.5 Sonnet cost?
At launch and currently (April 2026), Claude 3.5 Sonnet is priced at USD 3 per million input tokens and USD 15 per million output tokens — the same headline pricing as subsequent Sonnet tiers.
Sources
- Anthropic — Claude 3.5 Sonnet — accessed 2026-04-20
- Anthropic — Pricing — accessed 2026-04-20