# DeepSeek LLM 67B vs GPT-4.5

> GPT-4.5 is the stronger model overall, scoring 36.9 to 24.8 on the Noometry Index.

- Canonical page: https://noometry.com/compare/deepseek-llm-67b-vs-gpt-4-5
- Last updated: 2026-10-10
- Shared benchmarks: 14

## Summary

- They share 14 benchmarks with published results for both. DeepSeek LLM 67B scores higher in 1 category and GPT-4.5 in 7 categories; 8 gaps are clear of the uncertainty.
- The widest gap is in knowledge, where GPT-4.5 leads 32.4 to 7.1.
- The biggest single-benchmark swing is MATH Level 5: 6.4% for DeepSeek LLM 67B and 78.6% for GPT-4.5.
- DeepSeek LLM 67B has downloadable open weights; the other is API-only.

## Snapshot

| | DeepSeek LLM 67B | GPT-4.5 |
|---|---|---|
| Provider | DeepSeek | OpenAI |
| Noometry Index | 24.8 | 36.9 |
| Rank | 341 | 200 |
| Context | — | — |
| Input $/M | — | — |
| Output $/M | — | — |
| Weights | Open | Proprietary |

## Coding

- DeepSeek LLM 67B: 31.3 (#273)
- GPT-4.5: 41.8 (#102)

| Benchmark | DeepSeek LLM 67B | GPT-4.5 |
|---|---|---|
| LMArena Coding | 1096 | 1397 |
| Aider Polyglot | — | 44.9% |
| WeirdML | — | 39.4% |
| LiveBench Coding | — | 75.2% |

## Agentic & Tool Use

- DeepSeek LLM 67B: —
- GPT-4.5: 27.6 (#92)

| Benchmark | DeepSeek LLM 67B | GPT-4.5 |
|---|---|---|
| Cybench | — | 17.5% |

## Reasoning

- DeepSeek LLM 67B: 15.9 (#300)
- GPT-4.5: 14.1 (#320)

| Benchmark | DeepSeek LLM 67B | GPT-4.5 |
|---|---|---|
| LMArena Hard Prompts | 1070 | 1404 |
| Epoch Capabilities Index | 110.54 | 136.74 |
| ARC-AGI-2 | — | 0.8% |
| SimpleBench | — | 34.5% |
| ARC-AGI-1 | — | 10.3% |
| Chess Puzzles | 0% | — |
| EnigmaEval | — | 3.2% |
| LiveBench Reasoning | — | 71.1% |
| LiveBench Data Analysis | — | 64.3% |
| ForecastBench | — | 61.7 |
| LiveBench | — | 69% |

## Math

- DeepSeek LLM 67B: 8.1 (#318)
- GPT-4.5: 31.0 (#200)

| Benchmark | DeepSeek LLM 67B | GPT-4.5 |
|---|---|---|
| OTIS Mock AIME 2024-2025 | 0.8% | 37.8% |
| LMArena Math | 1108 | 1412 |
| MATH Level 5 | 6.4% | 78.6% |
| LiveBench Math | — | 69.3% |

## Knowledge

- DeepSeek LLM 67B: 7.1 (#308)
- GPT-4.5: 32.4 (#205)

| Benchmark | DeepSeek LLM 67B | GPT-4.5 |
|---|---|---|
| GPQA Diamond | 24.6% | 68.7% |
| Humanity's Last Exam | — | 5.4% |
| Confabulations | — | 13.6% |
| LMArena Expert | — | 1394 |

## Multimodal

- DeepSeek LLM 67B: —
- GPT-4.5: 37.0 (#68)

| Benchmark | DeepSeek LLM 67B | GPT-4.5 |
|---|---|---|
| LMArena Vision | — | 1195 |
| VPCT | — | 45% |

## Multilingual

- DeepSeek LLM 67B: 29.5 (#262)
- GPT-4.5: 52.7 (#78)

| Benchmark | DeepSeek LLM 67B | GPT-4.5 |
|---|---|---|
| LMArena Non-English | 1073 | 1413 |
| LMArena Chinese | 1132 | 1421 |
| LMArena French | — | 1417 |
| LMArena German | — | 1457 |
| LMArena Japanese | — | 1416 |
| LMArena Korean | — | 1392 |
| LMArena Russian | — | 1419 |

## Instruction Following

- DeepSeek LLM 67B: 55.7 (#272)
- GPT-4.5: 72.8 (#129)

| Benchmark | DeepSeek LLM 67B | GPT-4.5 |
|---|---|---|
| LMArena Instruction Following | 1079 | 1404 |
| LiveBench Instruction Following | — | 72.3% |

## Long Context

- DeepSeek LLM 67B: 33.1 (#260)
- GPT-4.5: 40.4 (#150)

| Benchmark | DeepSeek LLM 67B | GPT-4.5 |
|---|---|---|
| LMArena Longer Query | 1092 | 1406 |
| Fiction.LiveBench | — | 63.9% |

## Writing & Preference

- DeepSeek LLM 67B: 31.8 (#277)
- GPT-4.5: 57.1 (#129)

| Benchmark | DeepSeek LLM 67B | GPT-4.5 |
|---|---|---|
| LMArena Text | 1105 | 1417 |
| LMArena Creative Writing | 1067 | 1394 |
| LMArena Multi-Turn | 1082 | 1444 |
| Short-Story Creative Writing | — | 75.6% |
| EQ-Bench Creative Writing | — | 1258 |
| LiveBench Language | — | 61.5% |

## FAQ

### Is DeepSeek LLM 67B better than GPT-4.5?

GPT-4.5 is the stronger model overall, scoring 36.9 to 24.8 on the Noometry Index.

### Is DeepSeek LLM 67B or GPT-4.5 better for coding?

GPT-4.5 scores higher on coding benchmarks: 41.8 versus 31.3 in the Noometry coding category.

### How many benchmarks do DeepSeek LLM 67B and GPT-4.5 share?

14 benchmarks have published results for both models. DeepSeek LLM 67B has 15 scored results on Noometry and GPT-4.5 has 42.