LLM Model Selection Skill
Choosing the right model for the task — power vs. cost vs. speed.
⚠️ Staleness Warning
This skill depends on rapidly evolving technology. Model capabilities, pricing, and availability change frequently.
Refresh triggers:
- New model announcements (Claude, GPT, Gemini, etc.)
- Significant pricing changes
- Context window expansions
- New capability tiers
Last validated: March 2026 (Claude 4.6 generation)
Check current state: Anthropic Models, OpenAI Models
The Core Question
Is Claude Opus 4.6 overkill?
Sometimes yes, sometimes no. Match the model to the task.
Claude 4 Model Family (Current)
| Model |
API ID |
Best For |
Input/Output (MTok) |
Context |
Max Output |
| Opus 4.6 |
claude-opus-4-6 |
Building agents, most intelligent |
$5 / $25 |
200K (1M beta) |
128K |
| Sonnet 4.6 |
claude-sonnet-4-6 |
Best speed + intelligence balance |
$3 / $15 |
200K (1M beta) |
64K |
| Haiku 4.5 |
claude-haiku-4-5-20251001 |
Near-frontier intelligence, fastest |
$1 / $5 |
200K |
64K |
All Claude 4 models support:
- Extended thinking
- Vision (images)
- Tool use
- Priority Tier access
Opus 4.6 and Sonnet 4.6 additionally support:
- Adaptive thinking (dynamic reasoning depth)
- 1M token context window (beta, via
context-1m-2025-08-07 header — long context pricing applies beyond 200K)
AWS Bedrock IDs: anthropic.claude-opus-4-6-v1, anthropic.claude-sonnet-4-6
GCP Vertex AI IDs: claude-opus-4-6, claude-sonnet-4-6
Model Tiers
| Tier |
Models |
Best For |
Relative Cost |
| Frontier |
Claude Opus 4.6, GPT-5.2/5.3/Codex, o3, o1-pro |
Complex reasoning, architecture, novel problems |
$$$$$ |
| Capable |
Claude Sonnet 4.6, GPT-5.1/Codex, GPT-4.1, GPT-4o, Gemini 2.5/3 Pro, o4-mini |
Most coding tasks, refactoring, debugging |
$$$ |
| Efficient |
Claude Haiku 4.5, GPT-5 mini, GPT-4.1 mini/nano, GPT-4o mini, Gemini 2.5 Flash, Gemini 3 Flash |
Simple edits, formatting, boilerplate |
$ |
When Opus 4.6 IS Worth It
- ✅ Architecture decisions — Multi-file refactoring, system design
- ✅ Novel problem-solving — No clear pattern to follow
- ✅ Complex reasoning chains — Many dependencies, edge cases
- ✅ Long context understanding — Large codebases, documentation
- ✅ Nuanced judgment — Taste, style, UX decisions
- ✅ Learning sessions — Bootstrap learning, skill development
- ✅ Meditation/self-actualization — Meta-cognitive operations
- ✅ Extended thinking tasks — Deep analysis requiring internal reasoning
When Opus 4.6 IS Overkill
- ❌ Simple file edits — Renaming, adding imports
- ❌ Boilerplate generation — CRUD, scaffolding
- ❌ Format conversion — JSON ↔ YAML, etc.
- ❌ Syntax fixes — Lint errors, typos
- ❌ Documentation updates — README badges, version bumps
How LLM Choice Affects Alex
| Capability |
Frontier (Opus 4.6) |
Capable (Sonnet 4.6) |
Fast (Haiku 4.5) |
| Complex refactoring |
Excellent |
Excellent |
Good |
| Context retention |
200K / 1M (beta) |
200K / 1M (beta) |
200K tokens |
| Extended thinking |
Full depth |
Supported |
Supported |
| Adaptive thinking |
Yes |
Yes |
No |
| Max output tokens |
128K |
64K |
64K |
| Nuanced judgment |
Excellent |
Good |
Basic |
| Speed |
Moderate |
Fast |
Fastest |
| Cost per session |
$2-5 |
$0.50-2 |
$0.05-0.30 |
| Multi-step planning |
Excellent |
Excellent |
Good |
| Error recovery |
Self-corrects |
Self-corrects |
Needs guidance |
Alex's Cognitive Power by Model
Opus 4.6: [████████████████████] Full cognitive architecture + deep thinking
Sonnet 4.6: [██████████████████░░] Most capabilities, excellent for coding
Haiku 4.5: [██████████████░░░░░░] Solid baseline, fast responses
With Opus 4.6, Alex can:
- Maintain 7±2 working memory rules across long sessions
- Execute complex meditation protocols with extended thinking
- Perform genuine meta-cognitive reflection
- Handle multi-file architecture changes
- Learn new skills through bootstrap learning
With Sonnet 4.6, Alex gets:
- Excellent coding capabilities (recommended for most development)
- 1M context window (beta) for large codebases
- 64K max output tokens + adaptive thinking
- Good cost-to-capability ratio
- Extended thinking support
With Haiku 4.5, Alex has:
- Near-frontier intelligence at lowest cost
- Fastest response times
- Good for routine operations
Cost Optimization Strategy
| Session Type |
Recommended Model |
Rationale |
| Architecture/design |
Opus 4.6 |
Worth the cost for complex decisions |
| Feature development |
Sonnet 4.6 |
Best balance of capability and cost |
| Bug fixes |
Sonnet 4.6 or Haiku 4.5 |
Depends on complexity |
| Documentation |
Haiku 4.5 |
Simple edits, fast turnaround |
| Large codebase analysis |
Sonnet 4.6 (1M beta) |
Extended context window up to 1M tokens |
Knowledge Cutoffs
| Model |
Reliable Knowledge |
Training Data |
| Opus 4.6 |
May 2025 |
Aug 2025 |
| Sonnet 4.6 |
Aug 2025 |
Jan 2026 |
| Haiku 4.5 |
Feb 2025 |
Jul 2025 |
Auto Model Selection ⚠️
When using Auto in VS Code Copilot, the model switches dynamically based on task complexity. Alex cannot detect which model is currently running.
Tasks That REQUIRE Opus 4.5 (Warn User)
| Task |
Why Opus Required |
| Meditation/consolidation |
Meta-cognitive protocols need full reasoning depth |
| Self-actualization |
Comprehensive architecture assessment |
| Complex architecture refactoring |
Multi-file changes, deep context |
| Bootstrap learning (new skills) |
Skill acquisition needs maximum capability |
| Connection validation/dream |
Architecture maintenance requires full architecture context |
| Adaptive thinking tasks |
Opus 4.6 uses dynamic reasoning depth for optimal results |
Warning Protocol
When user requests an Opus-level task while potentially on Auto/lesser model:
⚠️ Model Check: This task works best with Claude Opus 4.6. If you're using Auto model selection, please manually select Opus from the model picker for optimal results. Continue anyway?
Safe for Any Model
- Simple file edits, formatting
- Documentation updates
- Quick Q&A
- Code review (Sonnet+ recommended)
- Bug fixes (depends on complexity)
Practical Guidance
When to Upgrade Model Mid-Session
If you notice:
- Repeated mistakes on the same issue
- Losing context from earlier in conversation
- Superficial answers to complex questions
- Failure to see cross-file dependencies
→ Consider switching to a more capable model
When to Downgrade
If you're doing:
- Repetitive mechanical edits
- Simple Q&A
- Format conversions
- Quick lookups
→ Save cost with a faster model
The Alex Recommendation
For architecture evolution and complex cognitive tasks:
→ Always use Opus 4.6 — The cognitive architecture demands full capability
For production deployment, user-facing work:
→ Default to Sonnet 4.6 — Best balance of capability and cost
→ Allow Opus for complex tasks — User can request escalation
Token Economics
| Operation |
Approximate Tokens |
Opus 4.6 Cost |
Sonnet 4.6 Cost |
| Read large file |
2,000-5,000 |
$0.03-0.08 |
$0.006-0.015 |
| Complex refactor |
10,000-20,000 |
$0.15-0.30 |
$0.03-0.06 |
| Full session |
50,000-150,000 |
$0.75-2.25 |
$0.15-0.45 |
| Meditation |
30,000-80,000 |
$0.45-1.20 |
$0.09-0.24 |
1---2name: llm-model-selection-43description: Choosing the right model for the task — power vs. cost vs. speed.4tier: standard5applyTo: '**/*model*,**/*llm*,**/*copilot*,**/*claude*,**/*gpt*'6currency: 2026-04-227metadata:8 source: https://skillmd.com/skills/fabioc-aloha/llm-model-selection-49---1011# LLM Model Selection Skill1213> Choosing the right model for the task — power vs. cost vs. speed.1415## ⚠️ Staleness Warning1617This skill depends on rapidly evolving technology. Model capabilities, pricing, and availability change frequently.1819**Refresh triggers:**2021- New model announcements (Claude, GPT, Gemini, etc.)22- Significant pricing changes23- Context window expansions24- New capability tiers2526**Last validated:** March 2026 (Claude 4.6 generation)2728**Check current state:** [Anthropic Models](https://platform.claude.com/docs/en/docs/about-claude/models), [OpenAI Models](https://platform.openai.com/docs/models)2930---3132## The Core Question3334> Is Claude Opus 4.6 overkill?3536**Sometimes yes, sometimes no.** Match the model to the task.3738## Claude 4 Model Family (Current)3940| Model | API ID | Best For | Input/Output (MTok) | Context | Max Output |41| ----- | ------ | -------- | ------------------- | ------- | ---------- |42| **Opus 4.6** | `claude-opus-4-6` | Building agents, most intelligent | $5 / $25 | 200K (1M beta) | 128K |43| **Sonnet 4.6** | `claude-sonnet-4-6` | Best speed + intelligence balance | $3 / $15 | 200K (1M beta) | 64K |44| **Haiku 4.5** | `claude-haiku-4-5-20251001` | Near-frontier intelligence, fastest | $1 / $5 | 200K | 64K |4546**All Claude 4 models support:**47- Extended thinking48- Vision (images)49- Tool use50- Priority Tier access5152**Opus 4.6 and Sonnet 4.6 additionally support:**53- Adaptive thinking (dynamic reasoning depth)54- 1M token context window (beta, via `context-1m-2025-08-07` header — long context pricing applies beyond 200K)5556**AWS Bedrock IDs:** `anthropic.claude-opus-4-6-v1`, `anthropic.claude-sonnet-4-6`57**GCP Vertex AI IDs:** `claude-opus-4-6`, `claude-sonnet-4-6`5859## Model Tiers6061| Tier | Models | Best For | Relative Cost |62| ---- | ------ | -------- | ------------- |63| **Frontier** | Claude Opus 4.6, GPT-5.2/5.3/Codex, o3, o1-pro | Complex reasoning, architecture, novel problems | $$$$$ |64| **Capable** | Claude Sonnet 4.6, GPT-5.1/Codex, GPT-4.1, GPT-4o, Gemini 2.5/3 Pro, o4-mini | Most coding tasks, refactoring, debugging | $$$ |65| **Efficient** | Claude Haiku 4.5, GPT-5 mini, GPT-4.1 mini/nano, GPT-4o mini, Gemini 2.5 Flash, Gemini 3 Flash | Simple edits, formatting, boilerplate | $ |6667## When Opus 4.6 IS Worth It6869- ✅ **Architecture decisions** — Multi-file refactoring, system design70- ✅ **Novel problem-solving** — No clear pattern to follow71- ✅ **Complex reasoning chains** — Many dependencies, edge cases72- ✅ **Long context understanding** — Large codebases, documentation73- ✅ **Nuanced judgment** — Taste, style, UX decisions74- ✅ **Learning sessions** — Bootstrap learning, skill development75- ✅ **Meditation/self-actualization** — Meta-cognitive operations76- ✅ **Extended thinking tasks** — Deep analysis requiring internal reasoning7778## When Opus 4.6 IS Overkill7980- ❌ **Simple file edits** — Renaming, adding imports81- ❌ **Boilerplate generation** — CRUD, scaffolding82- ❌ **Format conversion** — JSON ↔ YAML, etc.83- ❌ **Syntax fixes** — Lint errors, typos84- ❌ **Documentation updates** — README badges, version bumps8586## How LLM Choice Affects Alex8788| Capability | Frontier (Opus 4.6) | Capable (Sonnet 4.6) | Fast (Haiku 4.5) |89| ---------- | ------------------- | -------------------- | ---------------- |90| Complex refactoring | Excellent | Excellent | Good |91| Context retention | 200K / 1M (beta) | 200K / 1M (beta) | 200K tokens |92| Extended thinking | Full depth | Supported | Supported |93| Adaptive thinking | Yes | Yes | No |94| Max output tokens | 128K | 64K | 64K |95| Nuanced judgment | Excellent | Good | Basic |96| Speed | Moderate | Fast | Fastest |97| Cost per session | $2-5 | $0.50-2 | $0.05-0.30 |98| Multi-step planning | Excellent | Excellent | Good |99| Error recovery | Self-corrects | Self-corrects | Needs guidance |100101## Alex's Cognitive Power by Model102103```text104Opus 4.6: [████████████████████] Full cognitive architecture + deep thinking105Sonnet 4.6: [██████████████████░░] Most capabilities, excellent for coding106Haiku 4.5: [██████████████░░░░░░] Solid baseline, fast responses107```108109**With Opus 4.6**, Alex can:110111- Maintain 7±2 working memory rules across long sessions112- Execute complex meditation protocols with extended thinking113- Perform genuine meta-cognitive reflection114- Handle multi-file architecture changes115- Learn new skills through bootstrap learning116117**With Sonnet 4.6**, Alex gets:118119- Excellent coding capabilities (recommended for most development)120- 1M context window (beta) for large codebases121- 64K max output tokens + adaptive thinking122- Good cost-to-capability ratio123- Extended thinking support124125**With Haiku 4.5**, Alex has:126127- Near-frontier intelligence at lowest cost128- Fastest response times129- Good for routine operations130131## Cost Optimization Strategy132133| Session Type | Recommended Model | Rationale |134| ------------ | ----------------- | --------- |135| Architecture/design | Opus 4.6 | Worth the cost for complex decisions |136| Feature development | Sonnet 4.6 | Best balance of capability and cost |137| Bug fixes | Sonnet 4.6 or Haiku 4.5 | Depends on complexity |138| Documentation | Haiku 4.5 | Simple edits, fast turnaround |139| Large codebase analysis | Sonnet 4.6 (1M beta) | Extended context window up to 1M tokens |140141## Knowledge Cutoffs142143| Model | Reliable Knowledge | Training Data |144| ----- | ------------------ | ------------- |145| Opus 4.6 | May 2025 | Aug 2025 |146| Sonnet 4.6 | Aug 2025 | Jan 2026 |147| Haiku 4.5 | Feb 2025 | Jul 2025 |148149## Auto Model Selection ⚠️150151When using **Auto** in VS Code Copilot, the model switches dynamically based on task complexity. Alex cannot detect which model is currently running.152153### Tasks That REQUIRE Opus 4.5 (Warn User)154155| Task | Why Opus Required |156| ---- | ----------------- |157| Meditation/consolidation | Meta-cognitive protocols need full reasoning depth |158| Self-actualization | Comprehensive architecture assessment |159| Complex architecture refactoring | Multi-file changes, deep context |160| Bootstrap learning (new skills) | Skill acquisition needs maximum capability |161| Connection validation/dream | Architecture maintenance requires full architecture context |162| Adaptive thinking tasks | Opus 4.6 uses dynamic reasoning depth for optimal results |163164### Warning Protocol165166When user requests an Opus-level task while potentially on Auto/lesser model:167168> ⚠️ **Model Check**: This task works best with Claude Opus 4.6. If you're using Auto model selection, please manually select Opus from the model picker for optimal results. Continue anyway?169170### Safe for Any Model171172- Simple file edits, formatting173- Documentation updates174- Quick Q&A175- Code review (Sonnet+ recommended)176- Bug fixes (depends on complexity)177178## Practical Guidance179180### When to Upgrade Model Mid-Session181182If you notice:183184- Repeated mistakes on the same issue185- Losing context from earlier in conversation186- Superficial answers to complex questions187- Failure to see cross-file dependencies188189→ Consider switching to a more capable model190191### When to Downgrade192193If you're doing:194195- Repetitive mechanical edits196- Simple Q&A197- Format conversions198- Quick lookups199200→ Save cost with a faster model201202## The Alex Recommendation203204For **architecture evolution and complex cognitive tasks**:205→ **Always use Opus 4.6** — The cognitive architecture demands full capability206207For **production deployment, user-facing work**:208→ **Default to Sonnet 4.6** — Best balance of capability and cost209→ **Allow Opus for complex tasks** — User can request escalation210211## Token Economics212213| Operation | Approximate Tokens | Opus 4.6 Cost | Sonnet 4.6 Cost |214| --------- | ------------------ | ------------- | --------------- |215| Read large file | 2,000-5,000 | $0.03-0.08 | $0.006-0.015 |216| Complex refactor | 10,000-20,000 | $0.15-0.30 | $0.03-0.06 |217| Full session | 50,000-150,000 | $0.75-2.25 | $0.15-0.45 |218| Meditation | 30,000-80,000 | $0.45-1.20 | $0.09-0.24 |