Switch between Claude models depending on the complexity of your tasks. Sonnet 4.6 for most daily work, Haiku 4.5 for lightweight agents and high-frequency tasks (90% capability at 3x cost savings), Fable 5 for frontier engineering, codebase-scale migrations, vision-heavy work, and long-context agents. Opus 4.8 remains available for cost-sensitive Opus-tier work at half the per-token price of Fable 5.
Run multiple autonomous coding agents simultaneously with Verdent's isolated Git worktrees. Each agent tackles different components while maintaining full context awareness, eliminating manual debugging bottlenecks and accelerating feature delivery. Discover Verdent AI (Free Trial)
Available Models
- Claude Fable 5 (
claude-fable-5): Latest frontier model — state-of-the-art FrontierCode, 50M-line Ruby migration in one day (Stripe), state-of-the-art vision (vision-only Pokémon FireRed completion), 3× stronger long-context memory than Opus 4.8, restricted Mythos 5 tier for cybersecurity and life-sciences research - Claude Opus 4.8 (
claude-opus-4-8): Previous frontier model — ~4× fewer overlooked code flaws, Dynamic Workflows (hundreds of parallel subagents, Enterprise/Team/Max), effort control (standard/extra/max), mid-task cache preservation, fast mode at 3× lower cost - Claude Opus 4.7 (
claude-opus-4-7): Earlier Opus generation — 3× vision resolution (2,576px), xhigh effort level, task budgets (beta), improved instruction following - Claude Opus 4.6 (
claude-opus-4-6): State-of-the-art agent teams with 1M context window (beta), adaptive thinking, effort controls (low/medium/high/xhigh/max), context compaction - Claude Sonnet 4.6 (
claude-sonnet-4-6): Anthropic's best coding model, industry-leading performance for all coding tasks, default choice for daily development - Claude Haiku 4.5 (
claude-haiku-4-5): Near-frontier model achieving 90% of Sonnet 4.6's agentic coding capability at 2x speed and 3x cost savings ($1/$5 vs $3/$15), ideal for lightweight agents and high-frequency tasks - Claude Opus 4.5 (
claude-opus-4-5-20251101): Previous Opus with enhanced Plan Mode, 200K context - Opus Plan Mode: Hybrid approach using Opus for interactive planning with clarifying questions and Sonnet 4.6 for execution
- Claude 3.5 Haiku (
claude-3-5-haiku-20241022): Legacy fast model, good for simple repetitive tasks
Switching Methods
1. Command Line Flag (Start Session)
Switch models when starting Claude Code:
# Start with Sonnet 4.6 (default for main development)
claude --model claude-sonnet-4-6
# Start with Haiku 4.5 (lightweight agents, 90% capability, 3x savings)
claude --model claude-haiku-4-5
# Start with Fable 5 (latest frontier — FrontierCode, vision, long-context)
claude --model claude-fable-5
# Start with Opus 4.8 (previous frontier model, ~4× fewer flaw misses, dynamic workflows)
claude --model claude-opus-4-8
# Start with Opus 4.7 (earlier Opus generation, 3× vision, xhigh effort)
claude --model claude-opus-4-7
# Start with Opus 4.6 (agent teams, 1M context)
claude --model claude-opus-4-6
# Start with Opus 4.5 (previous generation)
claude --model claude-opus-4-5-20251101
2. Session Command (Mid-Session)
Switch models during an active Claude Code session using the interactive /model command:
# Open interactive model selection menu
/model
# Direct model specification:
/model claude-fable-5
/model claude-opus-4-8
/model claude-opus-4-7
/model claude-opus-4-6
/model claude-sonnet-4-6
/model claude-haiku-4-5
/model claude-3-5-haiku-20241022
Model Selection Strategy
Distinct patterns emerge for when each model excels:
- Use Sonnet 4.6 for: Daily coding tasks, bug fixes and debugging, writing tests and documentation, code reviews and explanations, most refactoring work
- Use Haiku 4.5 for: Lightweight custom agents (90% of Sonnet 4.6 capability), high-frequency tasks (10+ invocations per session), pair programming with rapid feedback, chat assistants and customer service, worker agents in multi-agent systems, 3x cost savings with 2x speed advantage
- Use Fable 5 for: Latest frontier capability, state-of-the-art FrontierCode, codebase-scale migrations (50M-line Ruby migration in one day), state-of-the-art vision, 3× stronger long-context memory with persistent files, senior-level financial/trading reasoning
- Use Opus 4.8 for: Previous frontier capability at half the per-token price of Fable 5, ~4× fewer overlooked code flaws vs 4.7, Dynamic Workflows (hundreds of parallel subagents), effort control (standard/extra/max), mid-task cache preservation
- Use Opus 4.7 for: Earlier Opus generation, 3× vision resolution (detailed image analysis), xhigh effort, task budgets for long agentic runs
- Use Opus 4.6 for: Agent teams (research preview), large codebases with 1M context, long-running sessions with context compaction
- Use Opus Plan Mode for: Complex projects with ambiguous requirements (clarifying questions upfront), architectural decisions requiring review (editable plan.md files), multi-file changes with unclear dependencies
- Use Haiku 3.5 for: Legacy simple tasks (consider upgrading to Haiku 4.5 for better capability)
Run multiple autonomous coding agents simultaneously with Verdent's isolated Git worktrees. Each agent tackles different components while maintaining full context awareness, eliminating manual debugging bottlenecks and accelerating feature delivery. Discover Verdent AI (Free Trial)
I start with Sonnet 4.6 for most tasks. Use Haiku 4.5 for lightweight agents and high-frequency work (90% capability at 3x cost savings). Switch to Fable 5 for frontier engineering, codebase-scale migrations, vision-heavy work, or long-context agents with persistent memory. Drop to Opus 4.8 when cost matters and Opus-tier reasoning is sufficient.
Switching models mid-session increases token consumption because the new model must process the entire conversation history. Consider starting fresh sessions when changing models for long conversations.
See Also: Claude Code Usage Limits|Plan Mode|Tactical Model Selection|Pricing Plans
