Claude 3.5 Sonnet: When the Mid-Tier Outperformed the Flagship
Claude 3.5 Sonnet launched June 2024 and outperformed the larger Claude 3 Opus on most benchmarks. A mid-tier model that redefined expectations.
Claude 3.5 Sonnet proved that bigger is not always better. Released June 20, 2024, this mid-tier model outperformed the larger, more expensive Claude 3 Opus on most benchmarks while costing 80% less. It was the moment developers stopped assuming that the flagship model was automatically the best choice.
Key Specs
| Spec | Details |
|---|---|
| API ID | claude-3-5-sonnet-20240620 |
| Context window | 200K tokens |
| Input pricing | $3 / 1M tokens |
| Output pricing | $15 / 1M tokens |
| Release date | June 20, 2024 |
| Max output tokens | 8,192 |
What Claude 3.5 Sonnet Brought to the Table
Flagship intelligence at mid-tier pricing. Claude 3.5 Sonnet matched or exceeded Claude 3 Opus across graduate-level reasoning (GPQA), undergraduate knowledge (MMLU), and coding (HumanEval). It did this at $3/$15 per million tokens instead of $15/$75. The cost-performance ratio was unprecedented.
Coding strength. This was the model that made Claude a serious coding tool. It solved 64% of problems on HumanEval, up from Opus's 55%. Developers who had been using GPT-4 for code started switching. The combination of reasoning quality and code output made it the default for software engineering tasks.
Speed. Claude 3.5 Sonnet operated at roughly twice the speed of Claude 3 Opus. For interactive coding sessions, chat applications, and any latency-sensitive workflow, the speed improvement was immediately noticeable.
The "Sonnet is enough" realization. Before this release, the assumption was simple: harder problem, bigger model. Claude 3.5 Sonnet broke that pattern. Teams that had been paying for Opus discovered they could get equal or better results from Sonnet at a fifth of the price. This shift in thinking influenced how developers approached model selection going forward.
How It Compared to Claude 3 Opus
| Benchmark | Claude 3 Opus | Claude 3.5 Sonnet |
|---|---|---|
| MMLU | 86.8% | 88.7% |
| GPQA | 50.4% | 59.4% |
| HumanEval | 55% | 64% |
| GSM8K | 95.0% | 96.4% |
Higher scores across the board, at one-fifth the cost. The numbers made the case on their own.
What About Claude 3.5 Opus?
Anthropic announced a Claude 3.5 Opus during the 3.5 Sonnet launch. It never shipped. The Claude 4 generation superseded the entire 3.5 line before a larger 3.5 model was needed. Given how well Sonnet performed, the market pressure for a 3.5 Opus simply never materialized.
Current Status
| Model | Status |
|---|---|
| Claude 3.5 Sonnet (v1) | Superseded by v2 (October 2024) |
The original 3.5 Sonnet was replaced by an upgraded version in October 2024 that brought further improvements and the groundbreaking Computer Use feature.
Navigation
- All Claude Models for the complete model index
- Claude 3, the predecessor family
- Claude 3.5 Sonnet v2, the October 2024 upgrade with Computer Use
Last updated on