Today, Giselle officially adds Anthropic's groundbreaking Claude 4 models—Opus 4 and Sonnet 4.
As someone deeply passionate about AI-powered coding, I've been relying on Claude 3.7 Sonnet for my development workflows. But with Claude 4's promise of being a leading coding model with state-of-the-art performance, I'm incredibly excited to explore what these new capabilities can unlock—and I'll be diving in to test them myself today.
Model-by-model highlights
Item | Opus 4 — ultimate coding powerhouse | Sonnet 4 — balanced performance champion |
---|---|---|
Ideal scenarios | Complex multi-step coding, sustained long-running tasks, advanced reasoning | Everyday coding, balanced performance, cost-effective workflows |
Flagship benchmarks | SWE-bench 72.5% / Terminal-bench 43.2% / Leading performance on coding benchmarks | SWE-bench 72.7% / Exceptional reasoning with superior efficiency |
Inference traits | Sustained performance for several hours; excels at complex, multi-file projects | High intelligence with optimal cost-performance balance |
Input | Text + Image | Text + Image |
Max context | 200k tokens | 200k tokens |
Output cap | 32k tokens | 64k tokens |
Which one to choose?
- Opus 4 – Choose when you need top-tier coding performance for complex refactoring, multi-repository analysis, or sustained agent workflows where capability and persistence are paramount.
- Sonnet 4 – Deploy for daily coding tasks, content generation, or any scenario where you need frontier-level performance without premium costs. Delivers exceptional value across most use cases.
Both models feature hybrid reasoning with two modes: instant responses for quick tasks and extended thinking for deeper analysis.
Why pair them with Giselle?
1. Hybrid Reasoning Modes
Claude 4 models automatically switch between standard and extended thinking modes based on task complexity. In Giselle, simply connect your data sources to a Claude Opus 4 or Sonnet 4 node, and the model will engage the appropriate reasoning depth—from instant responses to deep, multi-step analysis.
2. Advanced coding and tool use
Upload code repositories, documentation, or technical diagrams through File-Upload nodes and pipe them into Generation nodes. Claude 4 models excel at understanding complex codebases, suggesting improvements, and even maintaining context across multiple files for extended refactoring sessions.
3. Enhanced memory and context retention
With 200k-token context windows and improved memory capabilities, your agents can maintain awareness across extensive documents, remember key insights from previous interactions, and build tacit knowledge over time. Perfect for long-running analysis and iterative development workflows.
Note: Extended thinking mode and other Anthropic API features will be gradually implemented within Giselle. Currently, not all API features are supported.RetryClaude can make mistakes. Please double-check responses.
Get started now
- Log in — use GitHub, Google, or email.
- Choose a plan — Pro or higher unlocks Claude Opus 4 and Sonnet 4.
- Add a node — drag a claude-4-opus-20250514 or claude-4-sonnet-20250514 onto your studio.
- Connect input sources — link documents to the node.
- Write a prompt & run — the model automatically engages the right reasoning modefor optimal results.