Initiative
Plans before broad action
Useful when a change can affect several modules or decisions.
Claude models: Opus 4.6 vs Sonnet 4.6.
Use Claude Opus 4.6 for demanding reasoning and architecture; use Claude Sonnet 4.6 for coding and broad day-to-day agent tasks.
Floatboat · Claude
Current model
Claude
What do you want to finish?
Review this architecturePlan a repository-wide refactorTrace this failure across files
Describe the outcome or add project files…
- Live models
- Opus 4.6 · Sonnet 4.6
- Access
- Main client picker
- Configured next family
- Claude 5 · 1M / 128k budget
- API key
- Not required
Claude model characteristics
A deliberate collaborator for work with many dependencies.
Claude is strongest when the job benefits from understanding constraints before making broad changes. Opus leans into depth; Sonnet keeps that discipline with a faster everyday cadence.
Work span
Long-horizon
Maintains intent across research, implementation, and revision.
Best rhythm
Opus for depth, Sonnet for cadence
Choose within the family instead of treating both as interchangeable.
Watch for
Depth can add latency
Routine edits rarely need the most deliberate option.
Claude lineup
Pick depth or throughput without leaving the project.
The client exposes the 4.6 generation directly; newer families also appear in the broader Floatboat support matrix.
claude-opus-4-6
Complex architecture, long-horizon planning, and difficult reasoning.
Client picker
claude-sonnet-4-6
High-performance coding and general agent execution.
Client picker
claude-opus-5 / claude-sonnet-5 / claude-fable-5
Evaluation profiles with 1M context and 128k output budgets.
Not in the main picker
Work worth delegating
Use Claude where one missed dependency can undo the whole result.
These tasks reward coherence across steps rather than a fast first answer.
Repository-wide refactor
Task input
Change a shared subsystem without breaking existing callers.
Processing focus
Trace dependencies, define boundaries, change affected files, and verify the result.
Expected result: A reviewable multi-file change with its constraints preserved.
Architecture decision
Task input
Compare competing designs under real project constraints.
Processing focus
Keep tradeoffs, migration cost, and downstream impact in the same reasoning chain.
Expected result: A decision that explains both the choice and its consequences.
Cross-file failure trace
Task input
Find a failure whose symptom appears far from its cause.
Processing focus
Follow state and calls across modules before proposing a fix.
Expected result: A root-cause fix rather than a local patch over the symptom.
Claude in Floatboat
Reasoning connected to the project, not isolated in a chat box.
The value comes from pairing the selected Claude model with the files, rules, and execution flow already in Floatboat.
Complex reasoning
Opus 4.6 is the deliberate option for architecture and difficult multi-step decisions.
High-performance coding
Sonnet 4.6 is the everyday option for implementation and general agent work.
Long-running coherence
Keeps constraints in view while work moves from analysis to revision.
Managed access
Both live 4.6 models are selectable without a separate Anthropic API key.
Pick Claude by the cost of getting the reasoning wrong.
Choose Opus 4.6 when
- The task spans architecture, dependencies, and several rounds of execution.
- A weak early assumption would create expensive rework later.
- You need a careful technical plan before changing the project.
Choose Sonnet 4.6 or another route when
- The work is well-scoped and benefits from faster iteration.
- You are handling frequent everyday coding tasks.
- Very large mixed-media input is the main bottleneck; consider Gemini Pro.
Where Claude fits
Work that rewards coherence, not just a quick answer.
Architecture reviews
Hold constraints, tradeoffs, and dependencies together before proposing a design.
Repository-scale changes
Plan and inspect changes that cross modules instead of treating each file in isolation.
Long-running agent work
Maintain intent across research, implementation, verification, and revision.
Explore Other Model Families
Compare model families by the kind of work they support. Availability differs by model and entry point.
Claude · FAQ
When should I choose Opus instead of Sonnet?+
Choose Opus when reasoning depth and architectural consistency outweigh latency. Sonnet is the stronger default for everyday coding and agent work.
Is Claude Sonnet 4.6 good for coding?+
Yes. Sonnet 4.6 is positioned as the high-performance daily model for coding, implementation, and general agent work.
What kind of work benefits most from Claude?+
Claude is best suited to work that must preserve constraints across several steps, such as architecture decisions, repository-wide refactors, and cross-file failure analysis.
Is Claude Opus 4.6 better for complex reasoning?+
Opus 4.6 is the deeper option for architecture, difficult reasoning, and long-horizon work where consistency matters more than response speed.
What are Claude Opus 5, Sonnet 5, and Fable 5?+
They are evaluation and capacity profiles with configured 1M context and 128k output budgets. They are not current main-picker models.
Choose Opus depth or Sonnet pace for this project
Use Opus 4.6 for difficult architecture and Sonnet 4.6 for everyday coding or agent work, with the same project files in view.