Model comparison
DeepSeek-V3.2-Exp vs Trinity Large Thinking
DeepSeek-V3.2-Exp is the stronger model overall, scoring 44.9 to 38.1 on the Noometry Index.
Last verified . 23 shared benchmarks.
DeepSeek
Arcee AISummary
- They share 23 benchmarks with published results for both. DeepSeek-V3.2-Exp scores higher in 8 categories and Trinity Large Thinking in 0 categories; 8 gaps are clear of the uncertainty.
- The widest gap is in coding, where DeepSeek-V3.2-Exp leads 46.4 to 33.8.
- The biggest single-benchmark swing is Thematic Generalization: 65% for DeepSeek-V3.2-Exp and 41.6% for Trinity Large Thinking.
- DeepSeek-V3.2-Exp is cheaper at $0.26 / $0.38 per million input/output tokens, against $0.25 / $0.80 for Trinity Large Thinking.
- Trinity Large Thinking accepts more context: 262K tokens versus 164K.
Side by side
| DeepSeek-V3.2-Exp | Trinity Large Thinking | |
|---|---|---|
| Provider | DeepSeek | Arcee AI |
| Noometry Index | 44.9 | 38.1 |
| Released | 2025-09-29 | 2026-04-01 |
| Weights | Open | Open |
| Context window | 164K | 262K |
| Max output | 66K | 80K |
| Input $ / M tokens | $0.26 | $0.25 |
| Output $ / M tokens | $0.38 | $0.80 |
| Results tracked | 48 | 24 |
Sponsored placements are available on pages like this one. Advertise on Noometry
Category by category
Coding DeepSeek-V3.2-Exp leads
DeepSeek-V3.2-Exp: 46.4 (#61), Trinity Large Thinking: 33.8 (#238)
| Benchmark | DeepSeek-V3.2-Exp | Trinity Large Thinking |
|---|---|---|
| LMArena WebDev | 1362 | 1238 |
| SciCode | 38.9% | 36.1% |
| LMArena Coding | 1454 | 1381 |
| SWE-bench Verified (bash only) | 70% | — |
| Aider Polyglot | 74.2% | — |
| SWE-bench Multilingual | 59% | — |
| WeirdML | 39.5% | — |
Agentic & Tool Use Not comparable
DeepSeek-V3.2-Exp: 37.9 (#31), Trinity Large Thinking: —
| Benchmark | DeepSeek-V3.2-Exp | Trinity Large Thinking |
|---|---|---|
| Terminal-Bench | 39.6% | — |
| Berkeley Function Calling Leaderboard | 56.7% | — |
| TheAgentCompany | 42.9% | — |
| Vending-Bench 2 | 1,034 | — |
Reasoning DeepSeek-V3.2-Exp leads
DeepSeek-V3.2-Exp: 22.0 (#185), Trinity Large Thinking: 16.3 (#298)
| Benchmark | DeepSeek-V3.2-Exp | Trinity Large Thinking |
|---|---|---|
| NYT Connections (extended) | 36.7% | 16.5% |
| CritPt | 2.9% | 0.9% |
| Thematic Generalization | 65% | 41.6% |
| LMArena Hard Prompts | 1433 | 1350 |
| ARC-AGI-2 | 4% | — |
| Kagi LLM Benchmark | 52.2% | — |
| ARC-AGI-1 | 57% | — |
| Chess Puzzles | 14% | — |
| DTBench | 87.7% | — |
| LMCA | 29.1% | — |
| Surface Evolver Bench | — | 15.6% |
| Epoch Capabilities Index | 146.28 | — |
Math DeepSeek-V3.2-Exp leads
DeepSeek-V3.2-Exp: 40.8 (#75), Trinity Large Thinking: 34.9 (#149)
| Benchmark | DeepSeek-V3.2-Exp | Trinity Large Thinking |
|---|---|---|
| LMArena Math | 1435 | 1366 |
| MathArena Final-Answer Competitions | 57.7% | — |
| OTIS Mock AIME 2024-2025 | 87.8% | — |
| ProofBench | 8% | — |
| FrontierMath (Feb 2025 set) | 22.1% | — |
| FrontierMath Tier 4 (v1) | 2.1% | — |
Knowledge DeepSeek-V3.2-Exp leads
DeepSeek-V3.2-Exp: 51.6 (#61), Trinity Large Thinking: 40.3 (#107)
| Benchmark | DeepSeek-V3.2-Exp | Trinity Large Thinking |
|---|---|---|
| Vectara Hallucination Rate | 5.3% | 6.9% |
| LMArena Expert | 1436 | 1359 |
| GPQA Diamond | 83.4% | — |
Multilingual DeepSeek-V3.2-Exp leads
DeepSeek-V3.2-Exp: 52.4 (#85), Trinity Large Thinking: 46.4 (#155)
| Benchmark | DeepSeek-V3.2-Exp | Trinity Large Thinking |
|---|---|---|
| LMArena Non-English | 1409 | 1325 |
| LMArena Chinese | 1461 | 1372 |
| LMArena French | 1435 | 1374 |
| LMArena German | 1440 | 1356 |
| LMArena Japanese | 1377 | 1312 |
| LMArena Korean | 1372 | 1304 |
| LMArena Russian | 1423 | 1338 |
| LMArena Spanish | 1441 | 1356 |
Instruction Following DeepSeek-V3.2-Exp leads
DeepSeek-V3.2-Exp: 74.7 (#89), Trinity Large Thinking: 70.7 (#157)
| Benchmark | DeepSeek-V3.2-Exp | Trinity Large Thinking |
|---|---|---|
| LMArena Instruction Following | 1413 | 1334 |
Long Context DeepSeek-V3.2-Exp leads
DeepSeek-V3.2-Exp: 47.6 (#15), Trinity Large Thinking: 41.2 (#139)
| Benchmark | DeepSeek-V3.2-Exp | Trinity Large Thinking |
|---|---|---|
| LMArena Longer Query | 1428 | 1355 |
| Fiction.LiveBench | 83.3% | — |
| CL-bench | 13.2% | — |
| CL-bench Life | 9.5% | — |
Writing & Preference DeepSeek-V3.2-Exp leads
DeepSeek-V3.2-Exp: 62.5 (#73), Trinity Large Thinking: 54.0 (#152)
| Benchmark | DeepSeek-V3.2-Exp | Trinity Large Thinking |
|---|---|---|
| LMArena Text | 1425 | 1340 |
| LMArena Creative Writing | 1403 | 1321 |
| LMArena Multi-Turn | 1427 | 1342 |
| EQ-Bench Creative Writing | 1515 | — |
Frequently asked questions
Is DeepSeek-V3.2-Exp better than Trinity Large Thinking?
DeepSeek-V3.2-Exp is the stronger model overall, scoring 44.9 to 38.1 on the Noometry Index.
Which is cheaper, DeepSeek-V3.2-Exp or Trinity Large Thinking?
DeepSeek-V3.2-Exp is cheaper. It lists at $0.26 per million input tokens and $0.38 per million output tokens; Trinity Large Thinking lists at $0.25 and $0.80.
Is DeepSeek-V3.2-Exp or Trinity Large Thinking better for coding?
DeepSeek-V3.2-Exp scores higher on coding benchmarks: 46.4 versus 33.8 in the Noometry coding category.
Which has the bigger context window?
Trinity Large Thinking does, with 262K tokens against 164K.
How many benchmarks do DeepSeek-V3.2-Exp and Trinity Large Thinking share?
23 benchmarks have published results for both models. DeepSeek-V3.2-Exp has 48 scored results on Noometry and Trinity Large Thinking has 24.