Model comparison
DeepSeek-V3.2-Exp vs Llama 3-70B
DeepSeek-V3.2-Exp is the stronger model overall, scoring 44.3 to 28.8 on the Noometry Index.
Last verified . 22 shared benchmarks.
DeepSeek
MetaSummary
- They share 22 benchmarks with published results for both. DeepSeek-V3.2-Exp scores higher in 9 categories and Llama 3-70B in 0 categories; 9 gaps are clear of the uncertainty.
- The widest gap is in knowledge, where DeepSeek-V3.2-Exp leads 51.7 to 20.8.
- The biggest single-benchmark swing is OTIS Mock AIME 2024-2025: 87.8% for DeepSeek-V3.2-Exp and 4.3% for Llama 3-70B.
Side by side
| DeepSeek-V3.2-Exp | Llama 3-70B | |
|---|---|---|
| Provider | DeepSeek | Meta |
| Noometry Index | 44.3 | 28.8 |
| Released | 2025-09-29 | 2024-04-18 |
| Weights | Open | Open |
| Context window | 164K | — |
| Max output | 66K | — |
| Input $ / M tokens | $0.26 | — |
| Output $ / M tokens | $0.38 | — |
| Results tracked | 49 | 31 |
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: 46.5 (#65), Llama 3-70B: 35.8 (#218)
| Benchmark | DeepSeek-V3.2-Exp | Llama 3-70B |
|---|---|---|
| LMArena Coding | 1454 | 1206 |
| SWE-bench Verified (bash only) | 70% | — |
| Aider Polyglot | 74.2% | — |
| LMArena WebDev | 1362 | — |
| SWE-bench Multilingual | 59% | — |
| SciCode | 38.9% | — |
| WeirdML | 39.5% | — |
| BigCodeBench Instruct | — | 43.6% |
| BigCodeBench Complete | — | 54.5% |
| HumanEval+ | — | 72% |
| MBPP+ | — | 69% |
Agentic & Tool Use DeepSeek-V3.2-Exp leads
DeepSeek-V3.2-Exp: 32.7 (#59), Llama 3-70B: 21.1 (#139)
| Benchmark | DeepSeek-V3.2-Exp | Llama 3-70B |
|---|---|---|
| Terminal-Bench | 39.6% | — |
| APEX-Agents | 21.3% | — |
| Berkeley Function Calling Leaderboard | 56.7% | — |
| TheAgentCompany | 42.9% | — |
| Cybench | — | 5% |
| Vending-Bench 2 | 1,034 | — |
Reasoning DeepSeek-V3.2-Exp leads
DeepSeek-V3.2-Exp: 22.1 (#208), Llama 3-70B: 18.0 (#288)
| Benchmark | DeepSeek-V3.2-Exp | Llama 3-70B |
|---|---|---|
| Kagi LLM Benchmark | 52.2% | 35.1% |
| LMArena Hard Prompts | 1434 | 1195 |
| DTBench | 87.7% | 54.2% |
| Epoch Capabilities Index | 146.27 | 122.93 |
| ARC-AGI-2 | 4% | — |
| NYT Connections (extended) | 36.7% | — |
| ARC-AGI-1 | 57% | — |
| CritPt | 2.9% | — |
| Chess Puzzles | 14% | — |
| Thematic Generalization | 65% | — |
| LMCA | 29.1% | — |
| ForecastBench | — | 57.1 |
| WinoGrande | — | 83.5% |
Math DeepSeek-V3.2-Exp leads
DeepSeek-V3.2-Exp: 41.7 (#87), Llama 3-70B: 12.8 (#305)
| Benchmark | DeepSeek-V3.2-Exp | Llama 3-70B |
|---|---|---|
| OTIS Mock AIME 2024-2025 | 87.8% | 4.3% |
| LMArena Math | 1435 | 1218 |
| MathArena Final-Answer Competitions | 57.7% | — |
| ProofBench | 8% | — |
| MATH Level 5 | — | 22.6% |
| FrontierMath (Feb 2025 set) | 22.1% | — |
| FrontierMath Tier 4 (v1) | 2.1% | — |
Knowledge DeepSeek-V3.2-Exp leads
DeepSeek-V3.2-Exp: 51.7 (#66), Llama 3-70B: 20.8 (#277)
| Benchmark | DeepSeek-V3.2-Exp | Llama 3-70B |
|---|---|---|
| GPQA Diamond | 83.4% | 40.6% |
| LMArena Expert | 1436 | 1149 |
| Vectara Hallucination Rate | 5.3% | — |
| MMLU | — | 79.3% |
Multilingual DeepSeek-V3.2-Exp leads
DeepSeek-V3.2-Exp: 52.2 (#90), Llama 3-70B: 33.6 (#251)
| Benchmark | DeepSeek-V3.2-Exp | Llama 3-70B |
|---|---|---|
| LMArena Non-English | 1409 | 1142 |
| LMArena Chinese | 1461 | 1114 |
| LMArena French | 1433 | 1232 |
| LMArena German | 1440 | 1169 |
| LMArena Japanese | 1374 | 1017 |
| LMArena Korean | 1371 | 1017 |
| LMArena Russian | 1424 | 1159 |
| LMArena Spanish | 1440 | 1241 |
Instruction Following DeepSeek-V3.2-Exp leads
DeepSeek-V3.2-Exp: 74.5 (#93), Llama 3-70B: 62.5 (#238)
| Benchmark | DeepSeek-V3.2-Exp | Llama 3-70B |
|---|---|---|
| LMArena Instruction Following | 1413 | 1194 |
Long Context DeepSeek-V3.2-Exp leads
DeepSeek-V3.2-Exp: 47.6 (#16), Llama 3-70B: 35.6 (#240)
| Benchmark | DeepSeek-V3.2-Exp | Llama 3-70B |
|---|---|---|
| LMArena Longer Query | 1428 | 1174 |
| Fiction.LiveBench | 83.3% | — |
| CL-bench | 13.2% | — |
| CL-bench Life | 9.5% | — |
Writing & Preference DeepSeek-V3.2-Exp leads
DeepSeek-V3.2-Exp: 62.4 (#77), Llama 3-70B: 42.8 (#231)
| Benchmark | DeepSeek-V3.2-Exp | Llama 3-70B |
|---|---|---|
| LMArena Text | 1425 | 1221 |
| LMArena Creative Writing | 1403 | 1210 |
| LMArena Multi-Turn | 1427 | 1223 |
| EQ-Bench Creative Writing | 1515 | — |
Frequently asked questions
Is DeepSeek-V3.2-Exp better than Llama 3-70B?
DeepSeek-V3.2-Exp is the stronger model overall, scoring 44.3 to 28.8 on the Noometry Index.
Is DeepSeek-V3.2-Exp or Llama 3-70B better for coding?
DeepSeek-V3.2-Exp scores higher on coding benchmarks: 46.5 versus 35.8 in the Noometry coding category.
How many benchmarks do DeepSeek-V3.2-Exp and Llama 3-70B share?
22 benchmarks have published results for both models. DeepSeek-V3.2-Exp has 49 scored results on Noometry and Llama 3-70B has 31.