GPT-5.6 Luna vs Grok 4.5
GPT-5.6 Luna leads Grok 4.5 by 2.0 points on the Shogo Agent Index, but their ranges overlap, so treat it as close. GPT-5.6 Luna wins 11 and Grok 4.5 wins 9 of 20 shared benchmarks. GPT-5.6 Luna is cheaper for agent workloads.
Add more models to this comparison →Pillars
| Pillar | GPT-5.6 Luna | Grok 4.5 |
|---|---|---|
| Agentic coding | 59 | 54 |
| Long-horizon autonomy | 61 | 59 |
| Knowledge-work agents | 39 | 42 |
| Computer use | 40 | 36 |
| Security & context learning | 51 | 48 |
Shared benchmarks
| Benchmark | GPT-5.6 Luna | Grok 4.5 |
|---|---|---|
| APEX-Agentsindex | 43.0% | 56.2% |
| DeepSWEindex | 67.2% | 53.8% |
| Vending-Bench 2index | $4,095 | $3,887 |
| ARC-AGI | 88.0% | 87.2% |
| ARC-AGI-2 | 59.5% | 52.6% |
| Chess Puzzles | 43.0% | 35.0% |
| CritPt | 20.6% | 15.4% |
| DTBench | 89.1% | 96.5% |
| FrontierCode | 39.8% | 42.4% |
| FrontierMath Tier 4 | 61.0% | 24.4% |
| FrontierMath Tiers 1-3 | 82.1% | 57.2% |
| GPQA Diamond | 91.6% | 93.4% |
| LMCA | 48.5% | 45.2% |
| MirrorCode | 21.1% | 32.2% |
| OTIS Mock AIME | 98.3% | 97.8% |
| ProofBench | 60.0% | 31.0% |
| SciCode | 53.6% | 55.0% |
| SimpleBench | 46.8% | 70.0% |
| SimpleQA Verified | 41.0% | 48.3% |
| WebDev Arena | 1519 | 1555 |
Specs and price
| GPT-5.6 Luna | Grok 4.5 | |
|---|---|---|
| Agent Index | 51.4 (#34) | 49.4 (#36) |
| Agent $/M | $0.10 | $0.81 |
| Input / output $/M | $0.20 / $1.20 | $2.00 / $6.00 |
| Context | 1.1M | 500K |
| Released | Jul 9, 2026 | Jul 8, 2026 |
| Weights | Closed | Closed |
| API hosts | 5 | 4 |
Don’t pick one. Route.
Shogo can run GPT-5.6 Luna and Grok 4.5 side by side and send each step of an agent to the one that’s cheaper for it.
Try both in Shogo