Claude Fable 5 vs GPT-5.6 Luna
Claude Fable 5 leads GPT-5.6 Luna by 12.7 points on the Shogo Agent Index, but their ranges overlap, so treat it as close. Claude Fable 5 wins 21 and GPT-5.6 Luna wins 0 of 21 shared benchmarks. GPT-5.6 Luna is cheaper for agent workloads.
Add more models to this comparison →Pillars
| Pillar | Claude Fable 5 | GPT-5.6 Luna |
|---|---|---|
| Agentic coding | 74 | 59 |
| Long-horizon autonomy | 69 | 61 |
| Knowledge-work agents | 52 | 39 |
| Computer use | 55 | 40 |
| Security & context learning | 59 | 51 |
Shared benchmarks
| Benchmark | Claude Fable 5 | GPT-5.6 Luna |
|---|---|---|
| APEX-Agentsindex | 63.6% | 43.0% |
| DeepSWEindex | 69.9% | 67.2% |
| Vending-Bench 2index | $5,680 | $4,095 |
| ARC-AGI | 98.5% | 88.0% |
| ARC-AGI-2 | 89.2% | 59.5% |
| Chess Puzzles | 47.0% | 43.0% |
| CritPt | 28.6% | 20.6% |
| DTBench | 98.4% | 89.1% |
| FrontierCode | 53.5% | 39.8% |
| FrontierMath Tier 4 | 90.2% | 61.0% |
| FrontierMath Tiers 1-3 | 87.0% | 82.1% |
| GPQA Diamond | 93.9% | 91.6% |
| LMCA | 61.2% | 48.5% |
| MirrorCode | 70.0% | 21.1% |
| OTIS Mock AIME | 100.0% | 98.3% |
| ProofBench | 95.0% | 60.0% |
| SciCode | 61.0% | 53.6% |
| SimpleBench | 81.9% | 46.8% |
| SimpleQA Verified | 70.7% | 41.0% |
| WebDev Arena | 1628 | 1519 |
| WeirdML v3 | 15.6% | 5.2% |
Specs and price
| Claude Fable 5 | GPT-5.6 Luna | |
|---|---|---|
| Agent Index | 64.1 (#12) | 51.4 (#34) |
| Agent $/M | $4.62 | $0.10 |
| Input / output $/M | $10.00 / $50.00 | $0.20 / $1.20 |
| Context | 1M | 1.1M |
| Released | Jun 9, 2026 | Jul 9, 2026 |
| Weights | Closed | Closed |
| API hosts | 4 | 5 |
Don’t pick one. Route.
Shogo can run Claude Fable 5 and GPT-5.6 Luna side by side and send each step of an agent to the one that’s cheaper for it.
Try both in Shogo