Model comparison
DeepSeek-V3.2-Exp vs GPT-5.6 Terra
GPT-5.6 Terra is the stronger model overall, scoring 59.2 to 44.3 on the Noometry Index. DeepSeek-V3.2-Exp costs 16× less per token, which makes it the better buy when GPT-5.6 Terra's lead doesn't matter for your workload.
Last verified . 35 shared benchmarks.
DeepSeekSummary
- They share 35 benchmarks with published results for both. DeepSeek-V3.2-Exp scores higher in 1 category and GPT-5.6 Terra in 8 categories; 9 gaps are clear of the uncertainty.
- The widest gap is in math, where GPT-5.6 Terra leads 81.6 to 41.7.
- The biggest single-benchmark swing is ARC-AGI-2: 4% for DeepSeek-V3.2-Exp and 83.9% for GPT-5.6 Terra.
- DeepSeek-V3.2-Exp is cheaper at $0.26 / $0.38 per million input/output tokens, against $2 / $12 for GPT-5.6 Terra.
- GPT-5.6 Terra 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-5.6 Terra | |
|---|---|---|
| Provider | DeepSeek | OpenAI |
| Noometry Index | 44.3 | 59.2 |
| Released | 2025-09-29 | 2026-07-09 |
| Weights | Open | Proprietary |
| Context window | 164K | 1.05M |
| Max output | 66K | 128K |
| Input $ / M tokens | $0.26 | $2 |
| Output $ / M tokens | $0.38 | $12 |
| Results tracked | 49 | 52 |
Sponsored placements are available on pages like this one. Advertise on Noometry
Category by category
Coding GPT-5.6 Terra leads
DeepSeek-V3.2-Exp: 46.5 (#65), GPT-5.6 Terra: 57.7 (#19)
| Benchmark | DeepSeek-V3.2-Exp | GPT-5.6 Terra |
|---|---|---|
| LMArena WebDev | 1362 | 1522 |
| SciCode | 38.9% | 55% |
| WeirdML | 39.5% | 78.3% |
| LMArena Coding | 1454 | 1484 |
| DeepSWE | — | 69.6% |
| FrontierCode | — | 41.3% |
| SWE-bench Verified (bash only) | 70% | — |
| Aider Polyglot | 74.2% | — |
| CursorBench | — | 41.3% |
| SWE-bench Multilingual | 59% | — |
| ALE-Bench | — | 1,951 |
Agentic & Tool Use GPT-5.6 Terra leads
DeepSeek-V3.2-Exp: 32.7 (#59), GPT-5.6 Terra: 40.1 (#25)
| Benchmark | DeepSeek-V3.2-Exp | GPT-5.6 Terra |
|---|---|---|
| APEX-Agents | 21.3% | 58.2% |
| Vending-Bench 2 | 1,034 | 7,343 |
| Terminal-Bench | 39.6% | — |
| Berkeley Function Calling Leaderboard | 56.7% | — |
| TheAgentCompany | 42.9% | — |
| BALROG | — | 53.2% |
| GDP.pdf | — | 24.7% |
Reasoning GPT-5.6 Terra leads
DeepSeek-V3.2-Exp: 22.1 (#208), GPT-5.6 Terra: 60.7 (#21)
| Benchmark | DeepSeek-V3.2-Exp | GPT-5.6 Terra |
|---|---|---|
| ARC-AGI-2 | 4% | 83.9% |
| Kagi LLM Benchmark | 52.2% | 51.3% |
| NYT Connections (extended) | 36.7% | 78.4% |
| ARC-AGI-1 | 57% | 96.5% |
| CritPt | 2.9% | 30% |
| Chess Puzzles | 14% | 54% |
| LMArena Hard Prompts | 1434 | 1468 |
| DTBench | 87.7% | 93.3% |
| LMCA | 29.1% | 55% |
| Epoch Capabilities Index | 146.27 | 159.62 |
| SimpleBench | — | 48.9% |
| Thematic Generalization | 65% | — |
| Mystery Game Puzzles | — | 35% |
| Surface Evolver Bench | — | 83.8% |
Math GPT-5.6 Terra leads
DeepSeek-V3.2-Exp: 41.7 (#87), GPT-5.6 Terra: 81.6 (#12)
| Benchmark | DeepSeek-V3.2-Exp | GPT-5.6 Terra |
|---|---|---|
| OTIS Mock AIME 2024-2025 | 87.8% | 99.7% |
| ProofBench | 8% | 74% |
| LMArena Math | 1435 | 1466 |
| FrontierMath (Tiers 1-3) | — | 86% |
| FrontierMath Tier 4 | — | 70.7% |
| MathArena Final-Answer Competitions | 57.7% | — |
| FrontierMath (Feb 2025 set) | 22.1% | — |
| FrontierMath Tier 4 (v1) | 2.1% | — |
Knowledge GPT-5.6 Terra leads
DeepSeek-V3.2-Exp: 51.7 (#66), GPT-5.6 Terra: 61.2 (#30)
| Benchmark | DeepSeek-V3.2-Exp | GPT-5.6 Terra |
|---|---|---|
| GPQA Diamond | 83.4% | 93.3% |
| LMArena Expert | 1436 | 1492 |
| SimpleQA Verified | — | 43.2% |
| Vectara Hallucination Rate | 5.3% | — |
Multimodal Not comparable
DeepSeek-V3.2-Exp: —, GPT-5.6 Terra: 47.3 (#11)
| Benchmark | DeepSeek-V3.2-Exp | GPT-5.6 Terra |
|---|---|---|
| LMArena Vision | — | 1271 |
| Blueprint-Bench 2 | — | 30.8% |
| Furniture Assembly | — | 54.2% |
| LMArena Document | — | 1472 |
Multilingual GPT-5.6 Terra leads
DeepSeek-V3.2-Exp: 52.2 (#90), GPT-5.6 Terra: 54.4 (#44)
| Benchmark | DeepSeek-V3.2-Exp | GPT-5.6 Terra |
|---|---|---|
| LMArena Non-English | 1409 | 1439 |
| LMArena Chinese | 1461 | 1513 |
| LMArena French | 1433 | 1471 |
| LMArena German | 1440 | 1460 |
| LMArena Japanese | 1374 | 1457 |
| LMArena Korean | 1371 | 1425 |
| LMArena Russian | 1424 | 1450 |
| LMArena Spanish | 1440 | 1448 |
Instruction Following GPT-5.6 Terra leads
DeepSeek-V3.2-Exp: 74.5 (#93), GPT-5.6 Terra: 76.4 (#40)
| Benchmark | DeepSeek-V3.2-Exp | GPT-5.6 Terra |
|---|---|---|
| LMArena Instruction Following | 1413 | 1454 |
Long Context DeepSeek-V3.2-Exp leads
DeepSeek-V3.2-Exp: 47.6 (#16), GPT-5.6 Terra: 44.4 (#68)
| Benchmark | DeepSeek-V3.2-Exp | GPT-5.6 Terra |
|---|---|---|
| LMArena Longer Query | 1428 | 1451 |
| Fiction.LiveBench | 83.3% | — |
| CL-bench | 13.2% | — |
| CL-bench Life | 9.5% | — |
Writing & Preference GPT-5.6 Terra leads
DeepSeek-V3.2-Exp: 62.4 (#77), GPT-5.6 Terra: 70.2 (#23)
| Benchmark | DeepSeek-V3.2-Exp | GPT-5.6 Terra |
|---|---|---|
| LMArena Text | 1425 | 1447 |
| LMArena Creative Writing | 1403 | 1410 |
| EQ-Bench Creative Writing | 1515 | 1855 |
| LMArena Multi-Turn | 1427 | 1449 |
| EQ-Bench 4 | — | 1234 |
Frequently asked questions
Is DeepSeek-V3.2-Exp better than GPT-5.6 Terra?
GPT-5.6 Terra is the stronger model overall, scoring 59.2 to 44.3 on the Noometry Index. DeepSeek-V3.2-Exp costs 16× less per token, which makes it the better buy when GPT-5.6 Terra's lead doesn't matter for your workload.
Which is cheaper, DeepSeek-V3.2-Exp or GPT-5.6 Terra?
DeepSeek-V3.2-Exp is cheaper. It lists at $0.26 per million input tokens and $0.38 per million output tokens; GPT-5.6 Terra lists at $2 and $12.
Is DeepSeek-V3.2-Exp or GPT-5.6 Terra better for coding?
GPT-5.6 Terra scores higher on coding benchmarks: 57.7 versus 46.5 in the Noometry coding category.
Which has the bigger context window?
GPT-5.6 Terra does, with 1.05M tokens against 164K.
How many benchmarks do DeepSeek-V3.2-Exp and GPT-5.6 Terra share?
35 benchmarks have published results for both models. DeepSeek-V3.2-Exp has 49 scored results on Noometry and GPT-5.6 Terra has 52.