Model comparison
DeepSeek LLM 67B vs DeepSeek-V3
DeepSeek-V3 is the stronger model overall, scoring 39.5 to 24.9 on the Noometry Index.
Last verified . 14 shared benchmarks.
DeepSeekSummary
- They share 14 benchmarks with published results for both. DeepSeek LLM 67B scores higher in 0 categories and DeepSeek-V3 in 8 categories; 7 gaps are clear of the uncertainty.
- The widest gap is in knowledge, where DeepSeek-V3 leads 37.5 to 7.0.
- The biggest single-benchmark swing is MATH Level 5: 6.4% for DeepSeek LLM 67B and 75.5% for DeepSeek-V3.
Side by side
| DeepSeek LLM 67B | DeepSeek-V3 | |
|---|---|---|
| Provider | DeepSeek | DeepSeek |
| Noometry Index | 24.9 | 39.5 |
| Released | 2023-11-29 | 2024-12-26 |
| Weights | Open | Open |
| Context window | — | 164K |
| Max output | — | 164K |
| Input $ / M tokens | — | $0.24 |
| Output $ / M tokens | — | $0.90 |
| Results tracked | 15 | 60 |
Sponsored placements are available on pages like this one. Advertise on Noometry
Category by category
Coding DeepSeek-V3 leads
DeepSeek LLM 67B: 31.9 (#278), DeepSeek-V3: 42.3 (#106)
| Benchmark | DeepSeek LLM 67B | DeepSeek-V3 |
|---|---|---|
| LMArena Coding | 1096 | 1368 |
| Aider Polyglot | — | 55.1% |
| SciCode | — | 35.8% |
| WeirdML | — | 36.1% |
| BigCodeBench Instruct | — | 50% |
| LiveBench Coding | — | 70.9% |
| BigCodeBench Complete | — | 62.2% |
| HumanEval+ | — | 86.6% |
| MBPP+ | — | 73% |
Agentic & Tool Use Not comparable
DeepSeek LLM 67B: —, DeepSeek-V3: —
| Benchmark | DeepSeek LLM 67B | DeepSeek-V3 |
|---|---|---|
| METR Time Horizons | — | 49.6% |
Reasoning DeepSeek-V3 leads
DeepSeek LLM 67B: 16.5 (#304), DeepSeek-V3: 20.5 (#236)
| Benchmark | DeepSeek LLM 67B | DeepSeek-V3 |
|---|---|---|
| LMArena Hard Prompts | 1070 | 1365 |
| Epoch Capabilities Index | 110.5 | 135.94 |
| SimpleBench | — | 27.2% |
| Kagi LLM Benchmark | — | 52.3% |
| CritPt | — | 0% |
| Chess Puzzles | 0% | — |
| LiveBench Reasoning | — | 65.8% |
| DTBench | — | 64.8% |
| LiveBench Data Analysis | — | 60.9% |
| LMCA | — | 15.5% |
| BIG-Bench Hard | — | 87.5% |
| ForecastBench | — | 59.1 |
| HellaSwag | — | 88.9% |
| LiveBench | — | 66.9% |
| PIQA | — | 84.7% |
| WinoGrande | — | 85.2% |
Math DeepSeek-V3 leads
DeepSeek LLM 67B: 8.7 (#324), DeepSeek-V3: 32.1 (#219)
| Benchmark | DeepSeek LLM 67B | DeepSeek-V3 |
|---|---|---|
| OTIS Mock AIME 2024-2025 | 0.8% | 37.8% |
| LMArena Math | 1108 | 1373 |
| MATH Level 5 | 6.4% | 75.5% |
| Omni-MATH | — | 40.3% |
| LiveBench Math | — | 73.5% |
| FrontierMath (Feb 2025 set) | — | 1.7% |
Knowledge DeepSeek-V3 leads
DeepSeek LLM 67B: 7.0 (#313), DeepSeek-V3: 37.5 (#155)
| Benchmark | DeepSeek LLM 67B | DeepSeek-V3 |
|---|---|---|
| GPQA Diamond | 24.6% | 67.6% |
| MMLU-Pro | — | 72.3% |
| Confabulations | — | 26.1% |
| Vectara Hallucination Rate | — | 6.1% |
| GPQA (HELM) | — | 53.8% |
| LMArena Expert | — | 1351 |
| ARC (AI2) Challenge | — | 95.3% |
| MMLU | — | 87.2% |
| TriviaQA | — | 82.9% |
Multilingual DeepSeek-V3 leads
DeepSeek LLM 67B: 29.4 (#267), DeepSeek-V3: 48.5 (#143)
| Benchmark | DeepSeek LLM 67B | DeepSeek-V3 |
|---|---|---|
| LMArena Non-English | 1073 | 1358 |
| LMArena Chinese | 1132 | 1391 |
| LMArena French | — | 1385 |
| LMArena German | — | 1374 |
| LMArena Japanese | — | 1333 |
| LMArena Korean | — | 1319 |
| LMArena Russian | — | 1373 |
| LMArena Spanish | — | 1358 |
Instruction Following DeepSeek-V3 leads
DeepSeek LLM 67B: 55.4 (#277), DeepSeek-V3: 72.8 (#130)
| Benchmark | DeepSeek LLM 67B | DeepSeek-V3 |
|---|---|---|
| LMArena Instruction Following | 1079 | 1345 |
| LiveBench Instruction Following | — | 81.5% |
| IFEval | — | 83.2% |
Long Context Too close to call
DeepSeek LLM 67B: 33.1 (#265), DeepSeek-V3: 34.0 (#253)
| Benchmark | DeepSeek LLM 67B | DeepSeek-V3 |
|---|---|---|
| LMArena Longer Query | 1092 | 1352 |
| Fiction.LiveBench | — | 50% |
Writing & Preference DeepSeek-V3 leads
DeepSeek LLM 67B: 31.6 (#282), DeepSeek-V3: 57.4 (#130)
| Benchmark | DeepSeek LLM 67B | DeepSeek-V3 |
|---|---|---|
| LMArena Text | 1105 | 1375 |
| LMArena Creative Writing | 1067 | 1364 |
| LMArena Multi-Turn | 1082 | 1389 |
| Short-Story Creative Writing | — | 77% |
| EQ-Bench Creative Writing | — | 1472 |
| WildBench | — | 83% |
| LiveBench Language | — | 49.1% |
Frequently asked questions
Is DeepSeek LLM 67B better than DeepSeek-V3?
DeepSeek-V3 is the stronger model overall, scoring 39.5 to 24.9 on the Noometry Index.
Is DeepSeek LLM 67B or DeepSeek-V3 better for coding?
DeepSeek-V3 scores higher on coding benchmarks: 42.3 versus 31.9 in the Noometry coding category.
How many benchmarks do DeepSeek LLM 67B and DeepSeek-V3 share?
14 benchmarks have published results for both models. DeepSeek LLM 67B has 15 scored results on Noometry and DeepSeek-V3 has 60.