Claude 3.5 Sonnet Released: Faster, Better Coding 2024
Anthropic's Claude 3.5 Sonnet beats the previous Opus flagship on most benchmarks while running twice as fast and costing the same as the old mid-tier model. Here is what changed, why the price-performance shift matters, and how to decide whether to switch.
In this article
A New Generation, Not a Point Release
Claude 3.5 Sonnet is the first model in Anthropic's 3.5 generation. The naming is deliberate: it sits in the same Sonnet tier as the previous mid-range model, but the jump in capability is large enough that Anthropic positioned it above the older Opus flagship rather than as a sideways refresh. In practice, the model the company once reserved for its hardest tasks has been outclassed by the one it now recommends as a default.
What Changed
Claude 3.5 Sonnet delivers improvements across every major capability area Anthropic measures:
- 59% on GPQA (graduate-level science questions)
- Claude 3 Opus scored 50.4% on the same test
- Handles complex multi-step reasoning more reliably
- 64% on HumanEval, up from 38.4% for Claude 3 Sonnet
- 93% on Anthropic's internal agentic coding evaluation
- Stronger debugging and code explanation
- Improved chart and graph interpretation
- Better text extraction from images
- More accurate visual reasoning
- Roughly 2x faster than Claude 3 Opus
- Operates at the cost tier of the previous Claude 3 Sonnet
- No price increase over the model it replaces
Why This Matters
The usual pattern with AI models forces a tradeoff. Want the best output? Use the flagship and accept slower, pricier responses. Need speed and lower cost? Drop to a smaller model and accept weaker results. Teams building products end up routing requests between two or three models to balance the two.
Claude 3.5 Sonnet collapses that tradeoff. It is faster than the previous best model and stronger on most tasks, at a fraction of the price. For high-volume workloads, the economics are the real story: a team that was running Claude 3 Opus for quality can move to 3.5 Sonnet and cut per-request cost while improving output, rather than choosing between the two.
Coding Improvements
The coding jump is the most dramatic part of this release. Claude 3.5 Sonnet scored 64% on HumanEval, compared with 38.4% for Claude 3 Sonnet. That is not a marginal gain; it is the difference between a model you double-check constantly and one you can lean on for real work.
In practice that shows up as better code generation from plain-language descriptions, more accurate bug detection, clearer explanations of unfamiliar code, and more reliable refactoring suggestions. On Anthropic's internal agentic coding evaluation, where the model independently writes, debugs, and tests code with minimal hand-holding, it completed 93% of tasks. The takeaway for engineers is not that the model is autonomous, but that it can carry more of a multi-step coding task before a human needs to step in.
Artifacts: A New Way to Work With Output
Alongside the model, Anthropic introduced Artifacts in Claude.ai. When you ask Claude to generate something standalone, such as code, a document, a diagram, or a small web page, it opens in a dedicated panel beside the conversation instead of scrolling away inside the chat. You can edit and iterate on that output in place while the conversation continues on the side.
It is a small interface change with a real workflow effect. Artifacts turns Claude from a chat box that returns text into a workspace where the model's output is a living object you refine, which matters most for coding, drafting, and anything you build up over several turns.
Visual Analysis Upgrades
Claude 3.5 Sonnet handles visual inputs noticeably better than earlier versions. Anthropic calls it their strongest vision model to date, with the clearest gains on tasks that require reading detail out of an image:
- Transcribing text from imperfect or handwritten images
- Interpreting charts, graphs, and diagrams
- Extracting information from dense infographics
- Understanding visual context within an image
For business use, that translates into more reliable document analysis, better data extraction from screenshots and scanned files, and stronger interpretation of slide decks and reports, the kinds of tasks where a small misread can quietly break a downstream workflow.
Should You Switch?
For most workloads, yes, but the decision is worth a moment of thought:
- Running Claude 3 Opus today: test 3.5 Sonnet first, you will likely get better results, faster, for less.
- Using Claude 3 Sonnet or Haiku for cost reasons: 3.5 Sonnet is a clear upgrade at the same mid-tier price.
- Heavy coding or agentic workflows: this is the strongest reason to move; re-run your own evals to confirm on your codebase.
- Latency-critical, simple tasks: a smaller, cheaper model may still win, benchmark before assuming.
Test on your own data before you migrate. Public benchmarks tell you the model is stronger overall; they do not tell you how it behaves on your prompts, formats, and edge cases. Run a small side-by-side on real traffic before flipping production over. Claude API reference
Available Now
Claude 3.5 Sonnet is available immediately across Claude.ai (including the free tier), the Claude iOS app, the Anthropic API, Amazon Bedrock, and Google Cloud Vertex AI. Pricing matches the previous Claude 3 Sonnet tier, so existing integrations can switch by changing the model identifier.
What to Expect Next
This is the first release in the Claude 3.5 family. Anthropic has signaled that Claude 3.5 Haiku and Claude 3.5 Opus are planned to round out the lineup, which suggests the same generational leap is coming to both the fastest and the most capable tiers. If the Sonnet result is any indication, the next round will continue pushing intelligence-per-dollar rather than just raw capability.
Quick Takeaway
Claude 3.5 Sonnet delivers flagship-level performance at mid-tier speed and cost, and the new Artifacts panel makes its output easier to actually work with. The coding gains alone justify testing it for any technical workflow. If you are on Claude 3 Opus, the move to 3.5 Sonnet is close to a free upgrade: better results, faster, for less. Just validate it on your own tasks before you commit production traffic.
Claude models overview: compare Claude model capabilities and pricing across tiers. Learn more
Get the Claude playbook in your inbox.
One weekly email for Claude and Claude Code users. Real workflows, no hype. Subscribe and we send you The Claude Power-User Cheatsheet.
— ¶ —
Luke Thompson
Luke Thompson is the founder of The Operations Guide, LLC and editor of The Claude Insider. Based in Jonesborough, Tennessee, he has spent years building AI-augmented business systems and automation workflows for operators and teams. He began working with large language models in production well before the current wave of consumer AI tools, integrating them into client workflows, content pipelines, and operational infrastructure. At The Claude Insider, he writes about Claude with the specificity of someone who uses it daily as a professional tool — not as a reviewer or commentator, but as a builder. His coverage focuses on what actually works: prompt patterns, API integration strategies, agentic workflows, and the real-world tradeoffs that practitioners face. He is not affiliated with Anthropic, PBC.
Articles are researched and drafted with AI assistance, reviewed and edited by Luke Thompson.
Know where AI can pay off in your company.
Take the free two-minute AI Readiness Assessment. See your score, the two gaps holding you back, and the next move worth making.
