Claude 3.5 Sonnet Arrives, Beating Flagship Opus at Mid-Tier Pricing
On June 20, 2024, Anthropic announced “Claude 3.5 Sonnet.” Despite being positioned as a mid-tier model, it drew attention for outperforming the then-flagship Claude 3 Opus on many evaluations, including coding, while running at a significantly lower price and twice the speed of Opus.
Details
- Performance reversal: Scored 64% accuracy on an internal coding test versus Opus’s 38%. It also reportedly showed improvements in understanding nuance, humor, and complex instructions
- Speed and price: Twice as fast as Claude 3 Opus while keeping the mid-tier Sonnet pricing ($3 input / $15 output per million tokens)
- Stronger vision too: Surpassed Opus on vision benchmarks, with improved accuracy at transcribing text from blurry images
- New feature “Artifacts”: Launched alongside the model, this feature lets users view, edit, and collaborate on AI-generated content — such as code or website designs — in a dedicated window separate from the chat. It marked an attempt to expand Claude from a “conversational tool” into a “collaborative workspace”
- Availability: Free on Claude.ai and the iOS app, with relaxed rate limits for Pro/Team plans. Also available via the Anthropic API, Amazon Bedrock, and Google Cloud Vertex AI
What happened next
The pattern of “a mid-tier model surpassing the flagship” recurred repeatedly in later Claude releases. That October, an “upgraded Claude 3.5 Sonnet” with further improved performance arrived alongside the experimental “computer use” feature.