Model comparison
DeepSeek-V3.2-Exp vs GPT-6 Luna
GPT-6 Luna is the stronger model overall, scoring 53.3 to 44.3 on the Noometry Index.
Last verified . 31 shared benchmarks.
DeepSeekSummary
- They share 31 benchmarks with published results for both. DeepSeek-V3.2-Exp scores higher in 4 categories and GPT-6 Luna in 5 categories; 7 gaps are clear of the uncertainty.
- The widest gap is in math, where GPT-6 Luna leads 76.1 to 41.7.
- The biggest single-benchmark swing is ProofBench: 8% for DeepSeek-V3.2-Exp and 64% for GPT-6 Luna.
- GPT-6 Luna is cheaper at $0.10 / $0.50 per million input/output tokens, against $0.26 / $0.38 for DeepSeek-V3.2-Exp.
- GPT-6 Luna accepts more context: 1.05M tokens versus 164K.
- DeepSeek-V3.2-Exp has downloadable open weights; the other is API-only.
Side by side
| DeepSeek-V3.2-Exp | GPT-6 Luna | |
|---|---|---|
| Provider | DeepSeek | OpenAI |
| Noometry Index | 44.3 | 53.3 |
| Released | 2025-09-29 | 2026-09-22 |
| Weights | Open | Proprietary |
| Context window | 164K | 1.05M |
| Max output | 66K | 128K |
| Input $ / M tokens | $0.26 | $0.10 |
| Output $ / M tokens | $0.38 | $0.50 |
| Results tracked | 49 | 42 |
Sponsored placements are available on pages like this one. Advertise on Noometry
Category by category
Coding GPT-6 Luna leads
DeepSeek-V3.2-Exp: 46.5 (#65), GPT-6 Luna: 55.5 (#25)
| Benchmark | DeepSeek-V3.2-Exp | GPT-6 Luna |
|---|---|---|
| LMArena WebDev | 1362 | 1581 |
| SciCode | 38.9% | 54.6% |
| LMArena Coding | 1454 | 1439 |
| DeepSWE | — | 66.6% |
| FrontierCode | — | 42.4% |
| SWE-bench Verified (bash only) | 70% | — |
| Aider Polyglot | 74.2% | — |
| SWE-bench Multilingual | 59% | — |
| WeirdML | 39.5% | — |
| ALE-Bench | — | 1,577 |
Agentic & Tool Use Too close to call
DeepSeek-V3.2-Exp: 32.7 (#59), GPT-6 Luna: 33.3 (#54)
| Benchmark | DeepSeek-V3.2-Exp | GPT-6 Luna |
|---|---|---|
| APEX-Agents | 21.3% | 44.3% |
| Terminal-Bench | 39.6% | — |
| Berkeley Function Calling Leaderboard | 56.7% | — |
| TheAgentCompany | 42.9% | — |
| GDP.pdf | — | 23% |
| Vending-Bench 2 | 1,034 | — |
Reasoning GPT-6 Luna leads
DeepSeek-V3.2-Exp: 22.1 (#208), GPT-6 Luna: 48.2 (#41)
| Benchmark | DeepSeek-V3.2-Exp | GPT-6 Luna |
|---|---|---|
| ARC-AGI-2 | 4% | 59.3% |
| NYT Connections (extended) | 36.7% | 68.7% |
| ARC-AGI-1 | 57% | 86.7% |
| CritPt | 2.9% | 19.4% |
| Chess Puzzles | 14% | 31% |
| LMArena Hard Prompts | 1434 | 1411 |
| DTBench | 87.7% | 90.1% |
| LMCA | 29.1% | 44.5% |
| Epoch Capabilities Index | 146.27 | 156.28 |
| Kagi LLM Benchmark | 52.2% | — |
| Thematic Generalization | 65% | — |
| Mystery Game Puzzles | — | 7% |
Math GPT-6 Luna leads
DeepSeek-V3.2-Exp: 41.7 (#87), GPT-6 Luna: 76.1 (#15)
| Benchmark | DeepSeek-V3.2-Exp | GPT-6 Luna |
|---|---|---|
| OTIS Mock AIME 2024-2025 | 87.8% | 98.9% |
| ProofBench | 8% | 64% |
| LMArena Math | 1435 | 1416 |
| FrontierMath (Tiers 1-3) | — | 78.9% |
| FrontierMath Tier 4 | — | 56.1% |
| MathArena Final-Answer Competitions | 57.7% | — |
| FrontierMath (Feb 2025 set) | 22.1% | — |
| FrontierMath Tier 4 (v1) | 2.1% | — |
Knowledge GPT-6 Luna leads
DeepSeek-V3.2-Exp: 51.7 (#66), GPT-6 Luna: 57.0 (#41)
| Benchmark | DeepSeek-V3.2-Exp | GPT-6 Luna |
|---|---|---|
| GPQA Diamond | 83.4% | 90.5% |
| LMArena Expert | 1436 | 1444 |
| SimpleQA Verified | — | 41.4% |
| Vectara Hallucination Rate | 5.3% | — |
Multimodal Not comparable
DeepSeek-V3.2-Exp: —, GPT-6 Luna: 42.4 (#30)
| Benchmark | DeepSeek-V3.2-Exp | GPT-6 Luna |
|---|---|---|
| LMArena Vision | — | 1217 |
| Blueprint-Bench 2 | — | 31.2% |
| Furniture Assembly | — | 44.2% |
Multilingual DeepSeek-V3.2-Exp leads
DeepSeek-V3.2-Exp: 52.2 (#90), GPT-6 Luna: 50.5 (#117)
| Benchmark | DeepSeek-V3.2-Exp | GPT-6 Luna |
|---|---|---|
| LMArena Non-English | 1409 | 1386 |
| LMArena Chinese | 1461 | 1433 |
| LMArena French | 1433 | 1420 |
| LMArena German | 1440 | 1369 |
| LMArena Japanese | 1374 | 1369 |
| LMArena Korean | 1371 | 1360 |
| LMArena Russian | 1424 | 1394 |
| LMArena Spanish | 1440 | 1393 |
Instruction Following Too close to call
DeepSeek-V3.2-Exp: 74.5 (#93), GPT-6 Luna: 74.3 (#99)
| Benchmark | DeepSeek-V3.2-Exp | GPT-6 Luna |
|---|---|---|
| LMArena Instruction Following | 1413 | 1409 |
Long Context DeepSeek-V3.2-Exp leads
DeepSeek-V3.2-Exp: 47.6 (#16), GPT-6 Luna: 43.0 (#111)
| Benchmark | DeepSeek-V3.2-Exp | GPT-6 Luna |
|---|---|---|
| LMArena Longer Query | 1428 | 1409 |
| Fiction.LiveBench | 83.3% | — |
| CL-bench | 13.2% | — |
| CL-bench Life | 9.5% | — |
Writing & Preference DeepSeek-V3.2-Exp leads
DeepSeek-V3.2-Exp: 62.4 (#77), GPT-6 Luna: 58.3 (#119)
| Benchmark | DeepSeek-V3.2-Exp | GPT-6 Luna |
|---|---|---|
| LMArena Text | 1425 | 1391 |
| LMArena Creative Writing | 1403 | 1363 |
| LMArena Multi-Turn | 1427 | 1396 |
| EQ-Bench Creative Writing | 1515 | — |
Frequently asked questions
Is DeepSeek-V3.2-Exp better than GPT-6 Luna?
GPT-6 Luna is the stronger model overall, scoring 53.3 to 44.3 on the Noometry Index.
Which is cheaper, DeepSeek-V3.2-Exp or GPT-6 Luna?
GPT-6 Luna is cheaper. It lists at $0.10 per million input tokens and $0.50 per million output tokens; DeepSeek-V3.2-Exp lists at $0.26 and $0.38.
Is DeepSeek-V3.2-Exp or GPT-6 Luna better for coding?
GPT-6 Luna scores higher on coding benchmarks: 55.5 versus 46.5 in the Noometry coding category.
Which has the bigger context window?
GPT-6 Luna does, with 1.05M tokens against 164K.
How many benchmarks do DeepSeek-V3.2-Exp and GPT-6 Luna share?
31 benchmarks have published results for both models. DeepSeek-V3.2-Exp has 49 scored results on Noometry and GPT-6 Luna has 42.