# GPT-5.2 vs Trinity Large Thinking

> GPT-5.2 is the stronger model overall, scoring 54.1 to 38.6 on the Noometry Index. Trinity Large Thinking costs 12× less per token, which makes it the better buy when GPT-5.2's lead doesn't matter for your workload.

- Canonical page: https://noometry.com/compare/gpt-5-2-vs-trinity-large-thinking
- Last updated: 2026-10-10
- Shared benchmarks: 20

## Summary

- They share 20 benchmarks with published results for both. GPT-5.2 scores higher in 8 categories and Trinity Large Thinking in 0 categories; 8 gaps are clear of the uncertainty.
- The widest gap is in reasoning, where GPT-5.2 leads 50.2 to 16.9.
- The biggest single-benchmark swing is NYT Connections (extended): 83.6% for GPT-5.2 and 16.5% for Trinity Large Thinking.
- Trinity Large Thinking is cheaper at $0.25 / $0.80 per million input/output tokens, against $1.75 / $14 for GPT-5.2.
- GPT-5.2 accepts more context: 400K tokens versus 262K.
- Trinity Large Thinking has downloadable open weights; the other is API-only.

## Snapshot

| | GPT-5.2 | Trinity Large Thinking |
|---|---|---|
| Provider | OpenAI | Arcee AI |
| Noometry Index | 54.1 | 38.6 |
| Rank | 34 | 185 |
| Context | 400K | 262K |
| Input $/M | $1.75 | $0.25 |
| Output $/M | $14 | $0.80 |
| Weights | Proprietary | Open |

## Coding

- GPT-5.2: 51.6 (#37)
- Trinity Large Thinking: 34.1 (#244)

| Benchmark | GPT-5.2 | Trinity Large Thinking |
|---|---|---|
| LMArena WebDev | 1416 | 1238 |
| LMArena Coding | 1447 | 1381 |
| SWE-bench Verified | 73.8% | — |
| SWE-bench Verified (bash only) | 72.8% | — |
| SWE-bench Multilingual | 66.7% | — |
| SciCode | — | 36.1% |
| GSO | 27.4% | — |
| WeirdML | 72.2% | — |
| ALE-Bench | 1,294 | — |
| AlgoTune | 2.05 | — |

## Agentic & Tool Use

- GPT-5.2: 40.2 (#24)
- Trinity Large Thinking: —

| Benchmark | GPT-5.2 | Trinity Large Thinking |
|---|---|---|
| Terminal-Bench | 64.9% | — |
| Berkeley Function Calling Leaderboard | 55.9% | — |
| GDPval | 49.7% | — |
| Remote Labor Index | 2.5% | — |
| τ²-bench Airline | 83% | — |
| τ²-bench Banking | 32.2% | — |
| τ²-bench Retail | 81.6% | — |
| τ²-bench Telecom | 89.7% | — |
| DeepResearch Bench | 41.1% | — |
| LMArena Search | 1207 | — |
| METR Time Horizons | 75.3% | — |
| Vending-Bench 2 | 3,591 | — |

## Reasoning

- GPT-5.2: 50.2 (#35)
- Trinity Large Thinking: 16.9 (#298)

| Benchmark | GPT-5.2 | Trinity Large Thinking |
|---|---|---|
| NYT Connections (extended) | 83.6% | 16.5% |
| LMArena Hard Prompts | 1445 | 1350 |
| ARC-AGI-2 | 52.9% | — |
| SimpleBench | 45.8% | — |
| Kagi LLM Benchmark | 73.3% | — |
| ARC-AGI-1 | 86.2% | — |
| CritPt | — | 0.9% |
| Chess Puzzles | 49% | — |
| EnigmaEval | 10.4% | — |
| Thematic Generalization | — | 41.6% |
| EBR-Bench | 23% | — |
| Mystery Game Puzzles | 23% | — |
| DTBench | 90.9% | — |
| LMCA | 43.9% | — |
| Surface Evolver Bench | — | 15.6% |
| Epoch Capabilities Index | 153.45 | — |
| ForecastBench | 60.1 | — |

## Math

- GPT-5.2: 60.0 (#38)
- Trinity Large Thinking: 37.6 (#149)

| Benchmark | GPT-5.2 | Trinity Large Thinking |
|---|---|---|
| LMArena Math | 1440 | 1366 |
| FrontierMath (Tiers 1-3) | 67.4% | — |
| FrontierMath Tier 4 | 31.7% | — |
| MathArena Final-Answer Competitions | 72% | — |
| OTIS Mock AIME 2024-2025 | 96.1% | — |
| ProofBench | 15% | — |
| FrontierMath (Feb 2025 set) | 40.7% | — |
| FrontierMath Tier 4 (v1) | 18.8% | — |

## Knowledge

- GPT-5.2: 59.3 (#32)
- Trinity Large Thinking: 40.9 (#113)

| Benchmark | GPT-5.2 | Trinity Large Thinking |
|---|---|---|
| Vectara Hallucination Rate | 8.4% | 6.9% |
| LMArena Expert | 1445 | 1360 |
| GPQA Diamond | 91.4% | — |
| Humanity's Last Exam | 27.8% | — |
| SimpleQA Verified | 37.1% | — |

## Multimodal

- GPT-5.2: 51.3 (#7)
- Trinity Large Thinking: —

| Benchmark | GPT-5.2 | Trinity Large Thinking |
|---|---|---|
| LMArena Vision | 1268 | — |
| VPCT | 84% | — |
| Furniture Assembly | 38.3% | — |
| LMArena Document | 1405 | — |

## Multilingual

- GPT-5.2: 53.4 (#67)
- Trinity Large Thinking: 46.2 (#160)

| Benchmark | GPT-5.2 | Trinity Large Thinking |
|---|---|---|
| LMArena Non-English | 1425 | 1325 |
| LMArena Chinese | 1460 | 1373 |
| LMArena French | 1455 | 1374 |
| LMArena German | 1448 | 1356 |
| LMArena Japanese | 1420 | 1311 |
| LMArena Korean | 1392 | 1306 |
| LMArena Russian | 1440 | 1337 |
| LMArena Spanish | 1433 | 1357 |

## Instruction Following

- GPT-5.2: 74.7 (#89)
- Trinity Large Thinking: 70.5 (#162)

| Benchmark | GPT-5.2 | Trinity Large Thinking |
|---|---|---|
| LMArena Instruction Following | 1417 | 1334 |

## Long Context

- GPT-5.2: 44.0 (#78)
- Trinity Large Thinking: 41.3 (#144)

| Benchmark | GPT-5.2 | Trinity Large Thinking |
|---|---|---|
| LMArena Longer Query | 1428 | 1355 |
| CL-bench | 18.2% | — |

## Writing & Preference

- GPT-5.2: 66.8 (#32)
- Trinity Large Thinking: 53.8 (#158)

| Benchmark | GPT-5.2 | Trinity Large Thinking |
|---|---|---|
| LMArena Text | 1439 | 1340 |
| LMArena Creative Writing | 1401 | 1320 |
| LMArena Multi-Turn | 1458 | 1342 |
| EQ-Bench Creative Writing | 1703 | — |

## FAQ

### Is GPT-5.2 better than Trinity Large Thinking?

GPT-5.2 is the stronger model overall, scoring 54.1 to 38.6 on the Noometry Index. Trinity Large Thinking costs 12× less per token, which makes it the better buy when GPT-5.2's lead doesn't matter for your workload.

### Which is cheaper, GPT-5.2 or Trinity Large Thinking?

Trinity Large Thinking is cheaper. It lists at $0.25 per million input tokens and $0.80 per million output tokens; GPT-5.2 lists at $1.75 and $14.

### Is GPT-5.2 or Trinity Large Thinking better for coding?

GPT-5.2 scores higher on coding benchmarks: 51.6 versus 34.1 in the Noometry coding category.

### Which has the bigger context window?

GPT-5.2 does, with 400K tokens against 262K.

### How many benchmarks do GPT-5.2 and Trinity Large Thinking share?

20 benchmarks have published results for both models. GPT-5.2 has 67 scored results on Noometry and Trinity Large Thinking has 24.