Model comparison
DeepSeek V4.1 Flash vs GPT-5.6 Luna
GPT-5.6 Luna is the stronger model overall, scoring 54.6 to 52.8 on the Noometry Index. DeepSeek V4.1 Flash costs 1.7× less per token, which makes it the better buy when GPT-5.6 Luna's lead doesn't matter for your workload.
Last verified . 37 shared benchmarks.
DeepSeekSummary
- They share 37 benchmarks with published results for both. DeepSeek V4.1 Flash scores higher in 4 categories and GPT-5.6 Luna in 6 categories; 9 gaps are clear of the uncertainty.
- The widest gap is in math, where GPT-5.6 Luna leads 77.7 to 66.7.
- The biggest single-benchmark swing is FrontierMath Tier 4: 26.8% for DeepSeek V4.1 Flash and 61% for GPT-5.6 Luna.
- DeepSeek V4.1 Flash is cheaper at $0.15 / $0.60 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 1M.
- DeepSeek V4.1 Flash has downloadable open weights; the other is API-only.
Side by side
| DeepSeek V4.1 Flash | GPT-5.6 Luna | |
|---|---|---|
| Provider | DeepSeek | OpenAI |
| Noometry Index | 52.8 | 54.6 |
| Released | 2026-09-09 | 2026-07-09 |
| Weights | Open | Proprietary |
| Context window | 1M | 1.05M |
| Max output | 393K | 128K |
| Input $ / M tokens | $0.15 | $0.20 |
| Output $ / M tokens | $0.60 | $1.20 |
| Results tracked | 37 | 52 |
Sponsored placements are available on pages like this one. Advertise on Noometry
Category by category
Coding GPT-5.6 Luna leads
DeepSeek V4.1 Flash: 52.9 (#32), GPT-5.6 Luna: 54.5 (#28)
| Benchmark | DeepSeek V4.1 Flash | GPT-5.6 Luna |
|---|---|---|
| LMArena WebDev | 1619 | 1519 |
| SciCode | 51.9% | 53.6% |
| LMArena Coding | 1506 | 1466 |
| ALE-Bench | 1,092 | 1,667 |
| DeepSWE | — | 67.2% |
| FrontierCode | — | 39.8% |
| CursorBench | — | 35.9% |
| WeirdML | — | 60.9% |
Agentic & Tool Use GPT-5.6 Luna leads
DeepSeek V4.1 Flash: 31.2 (#69), GPT-5.6 Luna: 34.4 (#45)
| Benchmark | DeepSeek V4.1 Flash | GPT-5.6 Luna |
|---|---|---|
| APEX-Agents | 39.5% | 43% |
| GDP.pdf | 19.8% | 22.7% |
| BALROG | — | 45.6% |
| Vending-Bench 2 | — | 4,095 |
Reasoning DeepSeek V4.1 Flash leads
DeepSeek V4.1 Flash: 50.2 (#36), GPT-5.6 Luna: 47.6 (#43)
| Benchmark | DeepSeek V4.1 Flash | GPT-5.6 Luna |
|---|---|---|
| NYT Connections (extended) | 89.6% | 69.4% |
| CritPt | 14.3% | 20.6% |
| LMArena Hard Prompts | 1483 | 1451 |
| Mystery Game Puzzles | 43% | 21% |
| DTBench | 89.9% | 89.1% |
| LMCA | 47% | 48.5% |
| Surface Evolver Bench | 46.3% | 61.9% |
| Epoch Capabilities Index | 154.9 | 156.39 |
| ARC-AGI-2 | — | 59.5% |
| SimpleBench | — | 46.8% |
| Kagi LLM Benchmark | — | 49.1% |
| ARC-AGI-1 | — | 88% |
| Chess Puzzles | — | 40% |
Math GPT-5.6 Luna leads
DeepSeek V4.1 Flash: 66.7 (#25), GPT-5.6 Luna: 77.7 (#14)
| Benchmark | DeepSeek V4.1 Flash | GPT-5.6 Luna |
|---|---|---|
| FrontierMath (Tiers 1-3) | 67.4% | 82.1% |
| FrontierMath Tier 4 | 26.8% | 61% |
| OTIS Mock AIME 2024-2025 | 98.3% | 98.3% |
| ProofBench | 54% | 60% |
| LMArena Math | 1477 | 1458 |
Knowledge Too close to call
DeepSeek V4.1 Flash: 57.9 (#38), GPT-5.6 Luna: 58.5 (#34)
| Benchmark | DeepSeek V4.1 Flash | GPT-5.6 Luna |
|---|---|---|
| GPQA Diamond | 89.8% | 91.6% |
| LMArena Expert | 1506 | 1478 |
| SimpleQA Verified | — | 41% |
Multimodal GPT-5.6 Luna leads
DeepSeek V4.1 Flash: 39.1 (#61), GPT-5.6 Luna: 42.7 (#28)
| Benchmark | DeepSeek V4.1 Flash | GPT-5.6 Luna |
|---|---|---|
| LMArena Vision | 1277 | 1258 |
| Furniture Assembly | 34.2% | 42.5% |
| Blueprint-Bench 2 | — | 22.6% |
| LMArena Document | — | 1457 |
Multilingual DeepSeek V4.1 Flash leads
DeepSeek V4.1 Flash: 55.0 (#35), GPT-5.6 Luna: 52.8 (#78)
| Benchmark | DeepSeek V4.1 Flash | GPT-5.6 Luna |
|---|---|---|
| LMArena Non-English | 1448 | 1417 |
| LMArena Chinese | 1497 | 1470 |
| LMArena French | 1452 | 1456 |
| LMArena German | 1484 | 1454 |
| LMArena Japanese | 1412 | 1411 |
| LMArena Korean | 1452 | 1415 |
| LMArena Russian | 1471 | 1428 |
| LMArena Spanish | 1459 | 1448 |
Instruction Following DeepSeek V4.1 Flash leads
DeepSeek V4.1 Flash: 77.3 (#26), GPT-5.6 Luna: 75.6 (#57)
| Benchmark | DeepSeek V4.1 Flash | GPT-5.6 Luna |
|---|---|---|
| LMArena Instruction Following | 1474 | 1437 |
Long Context DeepSeek V4.1 Flash leads
DeepSeek V4.1 Flash: 45.2 (#47), GPT-5.6 Luna: 43.9 (#82)
| Benchmark | DeepSeek V4.1 Flash | GPT-5.6 Luna |
|---|---|---|
| LMArena Longer Query | 1475 | 1436 |
Writing & Preference GPT-5.6 Luna leads
DeepSeek V4.1 Flash: 65.4 (#48), GPT-5.6 Luna: 68.0 (#29)
| Benchmark | DeepSeek V4.1 Flash | GPT-5.6 Luna |
|---|---|---|
| LMArena Text | 1462 | 1431 |
| LMArena Creative Writing | 1435 | 1396 |
| EQ-Bench Creative Writing | 1540 | 1829 |
| LMArena Multi-Turn | 1457 | 1434 |
| EQ-Bench 4 | — | 1156 |
Frequently asked questions
Is DeepSeek V4.1 Flash better than GPT-5.6 Luna?
GPT-5.6 Luna is the stronger model overall, scoring 54.6 to 52.8 on the Noometry Index. DeepSeek V4.1 Flash costs 1.7× 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, DeepSeek V4.1 Flash or GPT-5.6 Luna?
DeepSeek V4.1 Flash is cheaper. It lists at $0.15 per million input tokens and $0.60 per million output tokens; GPT-5.6 Luna lists at $0.20 and $1.20.
Is DeepSeek V4.1 Flash or GPT-5.6 Luna better for coding?
GPT-5.6 Luna scores higher on coding benchmarks: 54.5 versus 52.9 in the Noometry coding category.
Which has the bigger context window?
GPT-5.6 Luna does, with 1.05M tokens against 1M.
How many benchmarks do DeepSeek V4.1 Flash and GPT-5.6 Luna share?
37 benchmarks have published results for both models. DeepSeek V4.1 Flash has 37 scored results on Noometry and GPT-5.6 Luna has 52.