Model comparison
Gemma 4 31B IT vs GPT-5.6 Luna
GPT-5.6 Luna is the stronger model overall, scoring 54.6 to 43.5 on the Noometry Index. Gemma 4 31B IT costs 3.0× less per token, which makes it the better buy when GPT-5.6 Luna's lead doesn't matter for your workload.
Last verified . 33 shared benchmarks.
GoogleSummary
- They share 33 benchmarks with published results for both. Gemma 4 31B IT scores higher in 2 categories and GPT-5.6 Luna in 7 categories; 7 gaps are clear of the uncertainty.
- The widest gap is in math, where GPT-5.6 Luna leads 77.7 to 43.2.
- The biggest single-benchmark swing is Chess Puzzles: 5% for Gemma 4 31B IT and 40% for GPT-5.6 Luna.
- Gemma 4 31B IT is cheaper at $0.09 / $0.34 per million input/output tokens, against $0.20 / $1.20 for GPT-5.6 Luna.
- GPT-5.6 Luna accepts more context: 1.05M tokens versus 262K.
- Gemma 4 31B IT has downloadable open weights; the other is API-only.
Side by side
| Gemma 4 31B IT | GPT-5.6 Luna | |
|---|---|---|
| Provider | OpenAI | |
| Noometry Index | 43.5 | 54.6 |
| Released | 2026-04-02 | 2026-07-09 |
| Weights | Open | Proprietary |
| Context window | 262K | 1.05M |
| Max output | 33K | 128K |
| Input $ / M tokens | $0.09 | $0.20 |
| Output $ / M tokens | $0.34 | $1.20 |
| Results tracked | 35 | 52 |
Sponsored placements are available on pages like this one. Advertise on Noometry
Category by category
Coding GPT-5.6 Luna leads
Gemma 4 31B IT: 42.3 (#108), GPT-5.6 Luna: 54.5 (#28)
| Benchmark | Gemma 4 31B IT | GPT-5.6 Luna |
|---|---|---|
| LMArena WebDev | 1366 | 1519 |
| SciCode | 43.4% | 53.6% |
| WeirdML | 52.3% | 60.9% |
| LMArena Coding | 1459 | 1466 |
| ALE-Bench | 925.5 | 1,667 |
| DeepSWE | — | 67.2% |
| FrontierCode | — | 39.8% |
| CursorBench | — | 35.9% |
Agentic & Tool Use Not comparable
Gemma 4 31B IT: —, GPT-5.6 Luna: 34.4 (#45)
| Benchmark | Gemma 4 31B IT | GPT-5.6 Luna |
|---|---|---|
| APEX-Agents | — | 43% |
| BALROG | — | 45.6% |
| GDP.pdf | — | 22.7% |
| Vending-Bench 2 | — | 4,095 |
Reasoning GPT-5.6 Luna leads
Gemma 4 31B IT: 27.2 (#122), GPT-5.6 Luna: 47.6 (#43)
| Benchmark | Gemma 4 31B IT | GPT-5.6 Luna |
|---|---|---|
| Kagi LLM Benchmark | 63.5% | 49.1% |
| NYT Connections (extended) | 70.6% | 69.4% |
| CritPt | 1.4% | 20.6% |
| Chess Puzzles | 5% | 40% |
| LMArena Hard Prompts | 1448 | 1451 |
| DTBench | 82.7% | 89.1% |
| LMCA | 39.3% | 48.5% |
| Surface Evolver Bench | 30.6% | 61.9% |
| Epoch Capabilities Index | 142.74 | 156.39 |
| ARC-AGI-2 | — | 59.5% |
| SimpleBench | — | 46.8% |
| ARC-AGI-1 | — | 88% |
| Thematic Generalization | 53% | — |
| Mystery Game Puzzles | — | 21% |
Math GPT-5.6 Luna leads
Gemma 4 31B IT: 43.2 (#81), GPT-5.6 Luna: 77.7 (#14)
| Benchmark | Gemma 4 31B IT | GPT-5.6 Luna |
|---|---|---|
| OTIS Mock AIME 2024-2025 | 73.3% | 98.3% |
| LMArena Math | 1465 | 1458 |
| FrontierMath (Tiers 1-3) | — | 82.1% |
| FrontierMath Tier 4 | — | 61% |
| ProofBench | — | 60% |
Knowledge GPT-5.6 Luna leads
Gemma 4 31B IT: 37.9 (#151), GPT-5.6 Luna: 58.5 (#34)
| Benchmark | Gemma 4 31B IT | GPT-5.6 Luna |
|---|---|---|
| GPQA Diamond | 75.8% | 91.6% |
| SimpleQA Verified | 10.4% | 41% |
| LMArena Expert | 1465 | 1478 |
| Vectara Hallucination Rate | 7.4% | — |
Multimodal GPT-5.6 Luna leads
Gemma 4 31B IT: 41.6 (#34), GPT-5.6 Luna: 42.7 (#28)
| Benchmark | Gemma 4 31B IT | GPT-5.6 Luna |
|---|---|---|
| LMArena Vision | 1277 | 1258 |
| LMArena Document | 1425 | 1457 |
| Blueprint-Bench 2 | — | 22.6% |
| Furniture Assembly | — | 42.5% |
Multilingual Gemma 4 31B IT leads
Gemma 4 31B IT: 53.8 (#57), GPT-5.6 Luna: 52.8 (#78)
| Benchmark | Gemma 4 31B IT | GPT-5.6 Luna |
|---|---|---|
| LMArena Non-English | 1431 | 1417 |
| LMArena Chinese | 1476 | 1470 |
| LMArena French | 1435 | 1456 |
| LMArena Russian | 1460 | 1428 |
| LMArena Spanish | 1444 | 1448 |
| LMArena German | — | 1454 |
| LMArena Japanese | — | 1411 |
| LMArena Korean | — | 1415 |
Instruction Following Too close to call
Gemma 4 31B IT: 75.5 (#61), GPT-5.6 Luna: 75.6 (#57)
| Benchmark | Gemma 4 31B IT | GPT-5.6 Luna |
|---|---|---|
| LMArena Instruction Following | 1433 | 1437 |
Long Context Too close to call
Gemma 4 31B IT: 44.2 (#71), GPT-5.6 Luna: 43.9 (#82)
| Benchmark | Gemma 4 31B IT | GPT-5.6 Luna |
|---|---|---|
| LMArena Longer Query | 1446 | 1436 |
Writing & Preference GPT-5.6 Luna leads
Gemma 4 31B IT: 60.5 (#96), GPT-5.6 Luna: 68.0 (#29)
| Benchmark | Gemma 4 31B IT | GPT-5.6 Luna |
|---|---|---|
| LMArena Text | 1443 | 1431 |
| LMArena Creative Writing | 1415 | 1396 |
| EQ-Bench Creative Writing | 1368 | 1829 |
| EQ-Bench 4 | 1120 | 1156 |
| LMArena Multi-Turn | 1452 | 1434 |
Frequently asked questions
Is Gemma 4 31B IT better than GPT-5.6 Luna?
GPT-5.6 Luna is the stronger model overall, scoring 54.6 to 43.5 on the Noometry Index. Gemma 4 31B IT costs 3.0× less per token, which makes it the better buy when GPT-5.6 Luna's lead doesn't matter for your workload.
Which is cheaper, Gemma 4 31B IT or GPT-5.6 Luna?
Gemma 4 31B IT is cheaper. It lists at $0.09 per million input tokens and $0.34 per million output tokens; GPT-5.6 Luna lists at $0.20 and $1.20.
Is Gemma 4 31B IT or GPT-5.6 Luna better for coding?
GPT-5.6 Luna scores higher on coding benchmarks: 54.5 versus 42.3 in the Noometry coding category.
Which has the bigger context window?
GPT-5.6 Luna does, with 1.05M tokens against 262K.
How many benchmarks do Gemma 4 31B IT and GPT-5.6 Luna share?
33 benchmarks have published results for both models. Gemma 4 31B IT has 35 scored results on Noometry and GPT-5.6 Luna has 52.