Model comparison
DeepSeek-V3.2-Exp vs Llama 3.1-8B
DeepSeek-V3.2-Exp is the stronger model overall, scoring 44.9 to 22.8 on the Noometry Index. Llama 3.1-8B costs 5.0× less per token, which makes it the better buy when DeepSeek-V3.2-Exp's lead doesn't matter for your workload.
Last verified . 28 shared benchmarks.
DeepSeek
MetaSummary
- They share 28 benchmarks with published results for both. DeepSeek-V3.2-Exp scores higher in 9 categories and Llama 3.1-8B in 0 categories; 9 gaps are clear of the uncertainty.
- The widest gap is in knowledge, where DeepSeek-V3.2-Exp leads 51.6 to 8.0.
- The biggest single-benchmark swing is OTIS Mock AIME 2024-2025: 87.8% for DeepSeek-V3.2-Exp and 1.7% for Llama 3.1-8B.
- Llama 3.1-8B is cheaper at $0.05 / $0.08 per million input/output tokens, against $0.26 / $0.38 for DeepSeek-V3.2-Exp.
- DeepSeek-V3.2-Exp accepts more context: 164K tokens versus 128K.
Side by side
| DeepSeek-V3.2-Exp | Llama 3.1-8B | |
|---|---|---|
| Provider | DeepSeek | Meta |
| Noometry Index | 44.9 | 22.8 |
| Released | 2025-09-29 | 2024-07-23 |
| Weights | Open | Open |
| Context window | 164K | 128K |
| Max output | 66K | 4K |
| Input $ / M tokens | $0.26 | $0.05 |
| Output $ / M tokens | $0.38 | $0.08 |
| Results tracked | 48 | 43 |
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), Llama 3.1-8B: 19.9 (#335)
| Benchmark | DeepSeek-V3.2-Exp | Llama 3.1-8B |
|---|---|---|
| SciCode | 38.9% | 13.2% |
| WeirdML | 39.5% | 1.7% |
| LMArena Coding | 1454 | 1195 |
| SWE-bench Verified (bash only) | 70% | — |
| Aider Polyglot | 74.2% | — |
| LMArena WebDev | 1362 | — |
| SWE-bench Multilingual | 59% | — |
| BigCodeBench Instruct | — | 32.8% |
| BigCodeBench Complete | — | 40.5% |
| HumanEval+ | — | 62.8% |
| MBPP+ | — | 55.6% |
Agentic & Tool Use DeepSeek-V3.2-Exp leads
DeepSeek-V3.2-Exp: 37.9 (#31), Llama 3.1-8B: 22.3 (#123)
| Benchmark | DeepSeek-V3.2-Exp | Llama 3.1-8B |
|---|---|---|
| Berkeley Function Calling Leaderboard | 56.7% | 25.8% |
| Terminal-Bench | 39.6% | — |
| TheAgentCompany | 42.9% | — |
| BALROG | — | 15.1% |
| Vending-Bench 2 | 1,034 | — |
Reasoning DeepSeek-V3.2-Exp leads
DeepSeek-V3.2-Exp: 22.0 (#185), Llama 3.1-8B: 14.5 (#316)
| Benchmark | DeepSeek-V3.2-Exp | Llama 3.1-8B |
|---|---|---|
| CritPt | 2.9% | 0% |
| Chess Puzzles | 14% | 0% |
| LMArena Hard Prompts | 1433 | 1175 |
| DTBench | 87.7% | 50.9% |
| LMCA | 29.1% | 5.4% |
| Epoch Capabilities Index | 146.28 | 116.48 |
| ARC-AGI-2 | 4% | — |
| Kagi LLM Benchmark | 52.2% | — |
| NYT Connections (extended) | 36.7% | — |
| ARC-AGI-1 | 57% | — |
| Thematic Generalization | 65% | — |
| PIQA | — | 81.2% |
Math DeepSeek-V3.2-Exp leads
DeepSeek-V3.2-Exp: 40.8 (#75), Llama 3.1-8B: 9.5 (#311)
| Benchmark | DeepSeek-V3.2-Exp | Llama 3.1-8B |
|---|---|---|
| OTIS Mock AIME 2024-2025 | 87.8% | 1.7% |
| LMArena Math | 1435 | 1179 |
| MathArena Final-Answer Competitions | 57.7% | — |
| ProofBench | 8% | — |
| Omni-MATH | — | 13.7% |
| MATH Level 5 | — | 22.9% |
| FrontierMath (Feb 2025 set) | 22.1% | — |
| FrontierMath Tier 4 (v1) | 2.1% | — |
| GSM8K | — | 82.4% |
Knowledge DeepSeek-V3.2-Exp leads
DeepSeek-V3.2-Exp: 51.6 (#61), Llama 3.1-8B: 8.0 (#302)
| Benchmark | DeepSeek-V3.2-Exp | Llama 3.1-8B |
|---|---|---|
| GPQA Diamond | 83.4% | 27% |
| LMArena Expert | 1436 | 1144 |
| MMLU-Pro | — | 40.6% |
| Vectara Hallucination Rate | 5.3% | — |
| GPQA (HELM) | — | 24.7% |
| BoolQ | — | 82.8% |
| MMLU | — | 56.1% |
Multilingual DeepSeek-V3.2-Exp leads
DeepSeek-V3.2-Exp: 52.4 (#85), Llama 3.1-8B: 34.2 (#244)
| Benchmark | DeepSeek-V3.2-Exp | Llama 3.1-8B |
|---|---|---|
| LMArena Non-English | 1409 | 1148 |
| LMArena Chinese | 1461 | 1151 |
| LMArena French | 1435 | 1176 |
| LMArena German | 1440 | 1144 |
| LMArena Japanese | 1377 | 1061 |
| LMArena Korean | 1372 | 1053 |
| LMArena Russian | 1423 | 1158 |
| LMArena Spanish | 1441 | 1169 |
Instruction Following DeepSeek-V3.2-Exp leads
DeepSeek-V3.2-Exp: 74.7 (#89), Llama 3.1-8B: 59.1 (#253)
| Benchmark | DeepSeek-V3.2-Exp | Llama 3.1-8B |
|---|---|---|
| LMArena Instruction Following | 1413 | 1159 |
| IFEval | — | 74.3% |
Long Context DeepSeek-V3.2-Exp leads
DeepSeek-V3.2-Exp: 47.6 (#15), Llama 3.1-8B: 35.8 (#233)
| Benchmark | DeepSeek-V3.2-Exp | Llama 3.1-8B |
|---|---|---|
| LMArena Longer Query | 1428 | 1182 |
| 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), Llama 3.1-8B: 29.9 (#285)
| Benchmark | DeepSeek-V3.2-Exp | Llama 3.1-8B |
|---|---|---|
| LMArena Text | 1425 | 1187 |
| LMArena Creative Writing | 1403 | 1154 |
| EQ-Bench Creative Writing | 1515 | 713 |
| LMArena Multi-Turn | 1427 | 1172 |
| WildBench | — | 68.7% |
Frequently asked questions
Is DeepSeek-V3.2-Exp better than Llama 3.1-8B?
DeepSeek-V3.2-Exp is the stronger model overall, scoring 44.9 to 22.8 on the Noometry Index. Llama 3.1-8B costs 5.0× less per token, which makes it the better buy when DeepSeek-V3.2-Exp's lead doesn't matter for your workload.
Which is cheaper, DeepSeek-V3.2-Exp or Llama 3.1-8B?
Llama 3.1-8B is cheaper. It lists at $0.05 per million input tokens and $0.08 per million output tokens; DeepSeek-V3.2-Exp lists at $0.26 and $0.38.
Is DeepSeek-V3.2-Exp or Llama 3.1-8B better for coding?
DeepSeek-V3.2-Exp scores higher on coding benchmarks: 46.4 versus 19.9 in the Noometry coding category.
Which has the bigger context window?
DeepSeek-V3.2-Exp does, with 164K tokens against 128K.
How many benchmarks do DeepSeek-V3.2-Exp and Llama 3.1-8B share?
28 benchmarks have published results for both models. DeepSeek-V3.2-Exp has 48 scored results on Noometry and Llama 3.1-8B has 43.