Claude Fable 5 vs GPT-5.6 Luna
Select two models to compare their coding performance, pricing, and features.
Key Takeaways
| Metric | Claude Fable 5 | GPT-5.6 Luna |
|---|---|---|
| Coding Index | 114.18 | 104.63 |
| Coding tier | S | A |
| Input Price | $10/1M | $0.2/1M |
| Output Price | $50/1M | $1.2/1M |
| Context Window | 1M | 1.05M |
Coding Index
Coding Index · Reference baseline: 100
Integer index difference: 9.55
DeepSeek V4 Flash 0731 = 100. The index reflects two shared tests, not capability multiples or statistically established gaps.
Verdict
The models have different editorial tier assessments. The index is supplementary and cannot predict performance on every development task.
Capabilities & Use Cases Comparison
Claude Fable 5 Tier S
- •Complex architecture
- •Multi-file refactoring
- •Claude Code
- •Code review
- •Highest coding benchmark score
- •Adaptive reasoning with Opus fallback
- •Top-tier architecture design
- •Best for complex coding tasks
GPT-5.6 Luna Tier A
- •Real-time completions
- •High-volume inline suggestions
- •Cost-sensitive workflows
- •Copilot autocomplete
- •Fast, low-latency code completions
- •Strong quality at a lightweight price point
- •Great for high-volume inline suggestions
- •Excellent cost efficiency for everyday coding
Claude Fable 5
Benchmark evidence and ranking method · Claude Fable 5
Cross-checked
Terminal-Bench 4.0 · Claude Code · max: 44.55% · 95% CI ± 3.85%
Artificial Analysis · Claude Fable 5 (Adaptive Reasoning, Max Effort, Opus 4.8 Fallback)
Test configuration: max
Terminal-Bench 2.1: 84.64% · SciCode: 61.00%
LiveBench · 2026-06-25 · claude-fable-5-max-effort
Coding: 85.99% · Agentic Coding: 62.17%
ModelRank assigns editorial tiers using public coding evidence, not automatic index thresholds or statistically proven classes. Price, speed, context size and sponsorship do not affect tiers.
DeepSeek V4 Flash 0731 = 100. The index reflects two shared tests, not capability multiples or statistically established gaps.
Results reflect particular model configurations. Agent harnesses, reasoning effort, tools, budgets and tasks may differ. Close indices do not establish superiority. B describes a relative position in this selected group, not an inability to code.
Data checked: 2026-09-10. Snapshot date is the retrieval date, not necessarily the evaluation date.
GPT-5.6 Luna
Benchmark evidence and ranking method · GPT-5.6 Luna
Cross-checked
Terminal-Bench 4.0 · Codex · max: 17.27% · 95% CI ± 2.85%
Artificial Analysis · GPT-5.6 Luna (max)
Test configuration: max
Terminal-Bench 2.1: 80.90% · SciCode: 53.59%
LiveBench · 2026-06-25 · gpt-5.6-luna-max
Coding: 82.91% · Agentic Coding: 48.43%
ModelRank assigns editorial tiers using public coding evidence, not automatic index thresholds or statistically proven classes. Price, speed, context size and sponsorship do not affect tiers.
DeepSeek V4 Flash 0731 = 100. The index reflects two shared tests, not capability multiples or statistically established gaps.
Results reflect particular model configurations. Agent harnesses, reasoning effort, tools, budgets and tasks may differ. Close indices do not establish superiority. B describes a relative position in this selected group, not an inability to code.
Data checked: 2026-09-10. Snapshot date is the retrieval date, not necessarily the evaluation date.
Results reflect particular model configurations. Agent harnesses, reasoning effort, tools, budgets and tasks may differ. Close indices do not establish superiority. B describes a relative position in this selected group, not an inability to code.
API cost estimate
1M input + 200K output (USD):
Claude Fable 5: $20.0000
GPT-5.6 Luna: $0.4400
Estimates use the displayed channel and time band; DeepSeek uses peak rates. Cache discounts, tool fees, long-context price tiers and uncounted additional tokens are excluded.
Frequently Asked Questions
Which model is better for coding: Claude Fable 5 or GPT-5.6 Luna?
Claude Fable 5 has a higher reference index on the selected shared benchmark configuration than GPT-5.6 Luna (114.18 vs 104.63, a 9.55-point difference). This does not by itself establish general coding superiority. Editorial tiers: Claude Fable 5 = S; GPT-5.6 Luna = A. Tiers are separate editorial assessments and do not establish a reliable order within a tier.
How do Claude Fable 5 and GPT-5.6 Luna compare in API pricing and costs?
Claude Fable 5: $10/1M input, $50/1M output (OpenRouter quote); GPT-5.6 Luna: $0.2/1M input, $1.2/1M output (OpenRouter quote). 1M input + 200K output (USD): $20.0000 vs $0.4400.
What are the context window sizes of Claude Fable 5 and GPT-5.6 Luna?
Claude Fable 5 provides a 1,000,000 token context window, while GPT-5.6 Luna provides 1,050,000 tokens.