GPT-5.6 Luna vs Qwen3.8 Max Preview
Qwen3.8 Max Preview leads GPT-5.6 Luna by 2.2 points on the Shogo Agent Index, but their ranges overlap, so treat it as close. GPT-5.6 Luna wins 7 and Qwen3.8 Max Preview wins 8 of 15 shared benchmarks. Qwen3.8 Max Preview is cheaper for agent workloads.
Add more models to this comparison →Pillars
| Pillar | GPT-5.6 Luna | Qwen3.8 Max Preview |
|---|---|---|
| Agentic coding | 59 | 58 |
| Long-horizon autonomy | 61 | 63 |
| Knowledge-work agents | 39 | 46 |
| Computer use | 40 | 41 |
| Security & context learning | 51 | 52 |
Shared benchmarks
| Benchmark | GPT-5.6 Luna | Qwen3.8 Max Preview |
|---|---|---|
| APEX-Agentsindex | 43.0% | 63.3% |
| DeepSWEindex | 67.2% | 57.5% |
| Chess Puzzles | 43.0% | 46.0% |
| CritPt | 20.6% | 20.0% |
| DTBench | 89.1% | 92.0% |
| FrontierMath Tier 4 | 61.0% | 46.3% |
| FrontierMath Tiers 1-3 | 82.1% | 74.7% |
| GPQA Diamond | 91.6% | 92.7% |
| LMCA | 48.5% | 46.2% |
| MirrorCode | 21.1% | 52.2% |
| OTIS Mock AIME | 98.3% | 99.4% |
| ProofBench | 60.0% | 58.0% |
| SciCode | 53.6% | 53.2% |
| SimpleQA Verified | 41.0% | 45.8% |
| WebDev Arena | 1519 | 1671 |
Specs and price
| GPT-5.6 Luna | Qwen3.8 Max Preview | |
|---|---|---|
| Agent Index | 51.4 (#34) | 53.6 (#30) |
| Agent $/M | $0.10 | Free |
| Input / output $/M | $0.20 / $1.20 | Free / Free |
| Context | 1.1M | 1M |
| Released | Jul 9, 2026 | Jul 19, 2026 |
| Weights | Closed | Closed |
| API hosts | 5 | 1 |
Don’t pick one. Route.
Shogo can run GPT-5.6 Luna and Qwen3.8 Max Preview side by side and send each step of an agent to the one that’s cheaper for it.
Try both in Shogo