Gemini 3.8 Flash vs GPT-5.6 Luna
Gemini 3.8 Flash leads GPT-5.6 Luna by 4.8 points on the Shogo Agent Index, but their ranges overlap, so treat it as close. Gemini 3.8 Flash wins 18 and GPT-5.6 Luna wins 4 of 22 shared benchmarks. GPT-5.6 Luna is cheaper for agent workloads.
Add more models to this comparison →Pillars
| Pillar | Gemini 3.8 Flash | GPT-5.6 Luna |
|---|---|---|
| Agentic coding | 63 | 59 |
| Long-horizon autonomy | 64 | 61 |
| Knowledge-work agents | 47 | 39 |
| Computer use | 44 | 40 |
| Security & context learning | 53 | 51 |
Shared benchmarks
| Benchmark | Gemini 3.8 Flash | GPT-5.6 Luna |
|---|---|---|
| APEX-Agentsindex | 64.3% | 43.0% |
| DeepSWEindex | 73.8% | 67.2% |
| Vending-Bench 2index | $5,094 | $4,095 |
| CursorBenchindex | 39.6% | 35.9% |
| ARC-AGI | 98.5% | 88.0% |
| ARC-AGI-2 | 89.2% | 59.5% |
| Chess Puzzles | 58.0% | 43.0% |
| CritPt | 18.3% | 20.6% |
| DTBench | 95.7% | 89.1% |
| FrontierCode | 41.2% | 39.8% |
| FrontierMath Tier 4 | 22.0% | 61.0% |
| FrontierMath Tiers 1-3 | 68.4% | 82.1% |
| GPQA Diamond | 95.4% | 91.6% |
| LMCA | 52.9% | 48.5% |
| MirrorCode | 38.9% | 21.1% |
| OTIS Mock AIME | 98.9% | 98.3% |
| ProofBench | 48.0% | 60.0% |
| SciCode | 56.6% | 53.6% |
| SimpleBench | 82.4% | 46.8% |
| SimpleQA Verified | 69.7% | 41.0% |
| WebDev Arena | 1568 | 1519 |
| WeirdML v3 | 7.4% | 5.2% |
Specs and price
| Gemini 3.8 Flash | GPT-5.6 Luna | |
|---|---|---|
| Agent Index | 56.2 (#21) | 51.4 (#34) |
| Agent $/M | $0.35 | $0.10 |
| Input / output $/M | $0.75 / $3.75 | $0.20 / $1.20 |
| Context | 1M | 1.1M |
| Released | Sep 2, 2026 | Jul 9, 2026 |
| Weights | Closed | Closed |
| API hosts | 6 | 5 |
Don’t pick one. Route.
Shogo can run Gemini 3.8 Flash 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