Skip to main content
Model flags let you choose which AI model to use. You can select by tier (Fable, Opus, Sonnet, Haiku) or specify an exact model ID.

​
Cross-Runtime Model Mapping

The same tier flag selects each runtime’s equivalent model:
TierFlagClaude CodeCodex CLI
Fable--fable / --bestclaude-fable-5gpt-5.4 (fallback — no Fable)
High--opus / --highclaude-opus-4-8gpt-5.4
Mid--sonnet / --midclaude-sonnet-4-6gpt-5.3-codex
Low--haiku / --lowclaude-haiku-4-5gpt-5.4-mini
--opus (the default) gives you the highest Opus-tier model for whichever runtime is active. --fable/--best selects Claude Fable 5, Anthropic’s most capable model — it’s available on Anthropic providers only, and on other runtimes (Ollama, LM Studio, Codex) it falls back to that runtime’s highest tier with a one-line notice.

​
Tier Flags

​
Fable Tier (Fable 5)

​
--fable
flag
Most capable model - Claude Fable 5, Anthropic’s most powerful model (a tier above Opus). Anthropic providers only; not the default.
ai --fable task.md
ai --aws --fable
ai --apikey --best          # --best is a synonym for --fable
Equivalent: --bestDefault model: claude-fable-5 (10/10/50 per MTok, 1M context)Availability: Anthropic API, AWS Bedrock, Vertex AI, Azure, Vercel, and Claude Pro (subject to your subscription). On Ollama, LM Studio, and Codex it falls back to that runtime’s highest tier and prints a notice.Data retention: Fable 5 is a Mythos-class Covered Model. Anthropic retains inputs/outputs for 30 days with human review on every platform that offers it — direct API, Pro, Bedrock, Vertex, and Azure alike. This is a property of the model, so switching providers doesn’t avoid it; the only way to avoid it is to not use Fable (use --opus).Cloud enablement: Because of that retention, Bedrock (--aws) and Vertex (--vertex) need a one-time, account/project-level data-sharing enablement before they’ll serve Fable, or requests fail (400/403). --apikey and --azure need no step — their standard config already shares with Anthropic (same retention, just no manual enablement). See the AWS Bedrock and Vertex AI provider docs for exact steps.
Exception: zero-data-retention (ZDR) organizations must explicitly turn retention on for Fable on every platform — including the direct API (Console → Workspace → Privacy Controls) and Azure — since ZDR and Covered Models are mutually exclusive.
Use for:
  • The most demanding reasoning and long-horizon agentic work
  • Tasks where Opus isn’t quite enough
​
--best
flag
Alias for --fable
ai --best task.md

​
High Tier (Opus)

​
--opus
flag
Highest-tier model - Most capable, best for complex reasoning
ai --opus task.md
ai --aws --opus
ai --ollama --opus      # Not applicable to Ollama
Equivalent: --highDefault model: claude-opus-4-8Use for:
  • Complex architectural decisions
  • Difficult refactoring
  • Security analysis
  • High-stakes code generation
​
--high
flag
Alias for --opus
ai --high task.md

​
Mid Tier (Sonnet)

​
--sonnet
flag
Mid-tier model - Balanced capability and speed
ai --sonnet task.md
ai --aws --sonnet
Equivalent: --midDefault model: claude-sonnet-4-6Use for:
  • Most coding tasks
  • General development
  • Documentation
  • Code review
​
--mid
flag
Alias for --sonnet
ai --mid task.md

​
Low Tier (Haiku)

​
--haiku
flag
Lowest-tier model - Fastest and most cost-effective
ai --haiku task.md
ai --aws --haiku
ai --ollama --haiku     # Not applicable to Ollama
Equivalent: --lowDefault model: claude-haiku-4-5Use for:
  • Quick tests
  • Simple tasks
  • Cost-sensitive workloads
  • High-volume automation
​
--low
flag
Alias for --haiku
ai --low task.md

​
Custom Model Selection

​
--model
flag
Specify exact model ID - Use any model supported by your provider
ai --model claude-opus-4-8
ai --aws --model global.anthropic.claude-opus-4-8
ai --ollama --model qwen3-coder
ai --vercel --model openai/gpt-5.2-codex
ai --codex --model gpt-5.4
Format depends on provider:
  • AWS Bedrock: global.anthropic.claude-opus-4-8
  • Vertex AI: claude-opus-4-8
  • Anthropic API: claude-opus-4-8
  • Ollama: Model name from ollama list
  • Vercel: provider/model-name (e.g., openai/gpt-4)
Precedence: --model overrides tier flags (--opus, --sonnet, --haiku)

​
Model Defaults Per Provider

​
Claude Subscription (Pro/Max)

ai --pro                 # Uses your subscription's latest model
Claude Pro/Max doesn’t support tier selection - it always uses the latest available model from your subscription.

​
API Providers (AWS, Vertex, Anthropic, Azure)

TierDefault Model
Fable (--fable, --best)claude-fable-5
Opus (--opus, --high)claude-opus-4-8
Sonnet (--sonnet, --mid)claude-sonnet-4-6
Haiku (--haiku, --low)claude-haiku-4-5
Default tier: High (Opus) — Fable is opt-in via --fable/--best. 
ai --aws                 # Uses Opus
ai --aws --sonnet        # Uses Sonnet
ai --aws --haiku         # Uses Haiku

​
Local Providers (Ollama, LM Studio)

ai --ollama --model qwen3-coder          # Specify model name
ai --lmstudio --model mlx-qwen-32b       # Specify model name
Local providers don’t have default models - you must specify with --model or configure a default in ~/.ai-runner/secrets.sh.

​
Codex CLI

TierDefault Model
High (--opus, --high)gpt-5.4
Mid (--sonnet, --mid)gpt-5.3-codex
Low (--haiku, --low)gpt-5.4-mini
Default: gpt-5.4 (Codex native default). Tier flags override: --mid selects gpt-5.3-codex. --fable/--best has no Codex equivalent, so it falls back to the High tier (gpt-5.4) with a notice. 
ai --codex                   # Uses gpt-5.4 (native default)
ai --codex --mid             # Uses gpt-5.3-codex
ai --codex --haiku           # Uses gpt-5.4-mini
Override in ~/.ai-runner/secrets.sh: 
export CODEX_MODEL_HIGH="gpt-5.4"
export CODEX_MODEL_MID="gpt-5.3-codex"
export CODEX_MODEL_LOW="gpt-5.4-mini"

​
Vercel AI Gateway

ai --vercel --model openai/gpt-5.2-codex        # OpenAI
ai --vercel --model anthropic/claude-opus-4.8   # Anthropic
ai --vercel --model google/gemini-exp-2506      # Google
Vercel supports 100+ models. Use provider/model-name format. See Vercel AI Gateway docs for available models.

​
Configuration

​
View Default Models

ai-status    # Shows default model IDs for each tier

​
Override Model Defaults

Edit ~/.ai-runner/secrets.sh: 
# Override AWS Sonnet model
export CLAUDE_MODEL_SONNET_AWS="global.anthropic.claude-sonnet-4-6"

# Override small/fast model (for background operations)
export CLAUDE_SMALL_FAST_MODEL_AWS="us.anthropic.claude-haiku-4-5-20251001-v1:0"

# Override Vertex Opus model
export CLAUDE_MODEL_OPUS_VERTEX="claude-opus-4-8"

# Override Fable 5 (top tier, --fable/--best) per provider
export CLAUDE_MODEL_FABLE_ANTHROPIC="claude-fable-5"
export CLAUDE_MODEL_FABLE_AWS="global.anthropic.claude-fable-5"
export CLAUDE_MODEL_FABLE_VERTEX="claude-fable-5"
export CLAUDE_MODEL_FABLE_AZURE="claude-fable-5"
export CLAUDE_MODEL_FABLE_VERCEL="anthropic/claude-fable-5"
See ~/.ai-runner/models.sh for all available override variables.

​
Dual Model Configuration

Claude Code uses two models:
  1. Primary model (ANTHROPIC_MODEL) - Interactive work, selected by tier flags
  2. Small/fast model (ANTHROPIC_SMALL_FAST_MODEL) - Background operations (defaults to Haiku)
When you run: 
ai --aws --opus
  • Primary: Opus (claude-opus-4-8)
  • Background: Haiku (claude-haiku-4-5)
Both models show in ai-status output.

​
Examples

​
Tier Selection

# Most capable (Opus)
ai --aws --opus complex-refactor.md

# Balanced (Sonnet, default)
ai --aws task.md
ai --aws --sonnet task.md     # Explicit

# Fastest/cheapest (Haiku)
ai --aws --haiku quick-test.md

​
Custom Models

# Specific AWS Bedrock model
ai --aws --model us.anthropic.claude-opus-4-8

# Ollama local model
ai --ollama --model qwen3-coder

# Ollama cloud model (no GPU needed)
ai --ollama --model minimax-m2.5:cloud

# Vercel with OpenAI
ai --vercel --model openai/gpt-5.2-codex

# Vercel with xAI Grok
ai --vercel --model xai/grok-2

​
Shebang Scripts

#!/usr/bin/env -S ai --aws --opus
Complex task requiring the most capable model.

#!/usr/bin/env -S ai --ollama --model qwen3-coder
Local execution with specific model.

​
Resume with Different Model

# Start with Sonnet
ai --aws --sonnet

# Resume with Opus for complex reasoning
ai --aws --opus --resume

# Resume with Haiku for speed
ai --aws --haiku --resume

​
Cost Optimization

# Use Haiku for bulk operations
for file in *.md; do
  ai --aws --haiku --skip "$file" >> results.txt
done

# Use Opus only for critical review
ai --aws --opus --skip final-review.md

​
Model Recommendations

Task TypeRecommended TierReason
Complex refactoringOpusNeeds deep understanding
Architectural designOpusHigh-stakes decisions
Security analysisOpusAccuracy critical
General codingSonnetBalanced performance
DocumentationSonnetGood quality/speed ratio
Code reviewSonnetSufficient capability
Quick testsHaikuSpeed matters
Bulk automationHaikuCost-effective
Simple tasksHaikuOverhead not needed

​
Troubleshooting

Model IDs are provider-specific. Check your provider’s available models:
# AWS Bedrock
aws bedrock list-foundation-models --region us-west-2

# Ollama
ollama list

# Vertex AI
gcloud ai models list --region=us-central1
See provider documentation for correct model ID format.
Check for overrides in ~/.ai-runner/secrets.sh:
grep CLAUDE_MODEL ~/.ai-runner/secrets.sh
Verify active model in session:
ai-status
Tier flags work differently with local and Codex providers. For Ollama/LM Studio, you may need --model to specify a local model name. For Codex, tiers map to GPT models automatically.
ai --ollama --model qwen3-coder

Provider Flags

Switch between cloud providers

Provider Model Lists

Available models per provider