Model comparison
DeepSeek-V3.2-Exp vs Llama 4 Maverick
DeepSeek-V3.2-Exp is the stronger model overall, scoring 44.3 to 30.9 on the Noometry Index.
Last verified . 36 shared benchmarks.
DeepSeek
MetaSummary
- They share 36 benchmarks with published results for both. DeepSeek-V3.2-Exp scores higher in 9 categories and Llama 4 Maverick in 0 categories; 9 gaps are clear of the uncertainty.
- The widest gap is in writing & preference, where DeepSeek-V3.2-Exp leads 62.4 to 38.8.
- The biggest single-benchmark swing is OTIS Mock AIME 2024-2025: 87.8% for DeepSeek-V3.2-Exp and 20.6% for Llama 4 Maverick.
- Both cost about the same: $0.26 input and $0.38 output per million tokens.
- DeepSeek-V3.2-Exp accepts more context: 164K tokens versus 128K.
Side by side
| DeepSeek-V3.2-Exp | Llama 4 Maverick | |
|---|---|---|
| Provider | DeepSeek | Meta |
| Noometry Index | 44.3 | 30.9 |
| Released | 2025-09-29 | 2025-04-05 |
| Weights | Open | Open |
| Context window | 164K | 128K |
| Max output | 66K | 4K |
| Input $ / M tokens | $0.26 | $0.19 |
| Output $ / M tokens | $0.38 | $0.65 |
| Results tracked | 49 | 54 |
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 4 Maverick: 26.6 (#324)
| Benchmark | DeepSeek-V3.2-Exp | Llama 4 Maverick |
|---|---|---|
| SWE-bench Verified (bash only) | 70% | 21% |
| Aider Polyglot | 74.2% | 15.6% |
| SciCode | 38.9% | 33.1% |
| WeirdML | 39.5% | 24.5% |
| LMArena Coding | 1454 | 1302 |
| LMArena WebDev | 1362 | — |
| SWE-bench Multilingual | 59% | — |
| BigCodeBench Instruct | — | 49.7% |
| BigCodeBench Complete | — | 61.4% |
| ALE-Bench | — | 172.97 |
Agentic & Tool Use DeepSeek-V3.2-Exp leads
DeepSeek-V3.2-Exp: 32.7 (#59), Llama 4 Maverick: 28.2 (#91)
| Benchmark | DeepSeek-V3.2-Exp | Llama 4 Maverick |
|---|---|---|
| Berkeley Function Calling Leaderboard | 56.7% | 37.3% |
| Terminal-Bench | 39.6% | — |
| APEX-Agents | 21.3% | — |
| TheAgentCompany | 42.9% | — |
| Vending-Bench 2 | 1,034 | — |
Reasoning DeepSeek-V3.2-Exp leads
DeepSeek-V3.2-Exp: 22.1 (#208), Llama 4 Maverick: 10.1 (#342)
| Benchmark | DeepSeek-V3.2-Exp | Llama 4 Maverick |
|---|---|---|
| ARC-AGI-2 | 4% | 0% |
| Kagi LLM Benchmark | 52.2% | 55.9% |
| NYT Connections (extended) | 36.7% | 8% |
| ARC-AGI-1 | 57% | 4.4% |
| CritPt | 2.9% | 0% |
| LMArena Hard Prompts | 1434 | 1281 |
| DTBench | 87.7% | 61.9% |
| LMCA | 29.1% | 15.9% |
| Epoch Capabilities Index | 146.27 | 132.2 |
| SimpleBench | — | 27.7% |
| Chess Puzzles | 14% | — |
| EnigmaEval | — | 0.6% |
| Thematic Generalization | 65% | — |
| ForecastBench | — | 57.5 |
Math DeepSeek-V3.2-Exp leads
DeepSeek-V3.2-Exp: 41.7 (#87), Llama 4 Maverick: 26.0 (#262)
| Benchmark | DeepSeek-V3.2-Exp | Llama 4 Maverick |
|---|---|---|
| OTIS Mock AIME 2024-2025 | 87.8% | 20.6% |
| LMArena Math | 1435 | 1299 |
| FrontierMath (Feb 2025 set) | 22.1% | 0.7% |
| MathArena Final-Answer Competitions | 57.7% | — |
| ProofBench | 8% | — |
| Omni-MATH | — | 42.2% |
| MATH Level 5 | — | 73% |
| FrontierMath Tier 4 (v1) | 2.1% | — |
Knowledge DeepSeek-V3.2-Exp leads
DeepSeek-V3.2-Exp: 51.7 (#66), Llama 4 Maverick: 33.4 (#204)
| Benchmark | DeepSeek-V3.2-Exp | Llama 4 Maverick |
|---|---|---|
| GPQA Diamond | 83.4% | 67% |
| Vectara Hallucination Rate | 5.3% | 8.2% |
| LMArena Expert | 1436 | 1259 |
| Humanity's Last Exam | — | 5.7% |
| MMLU-Pro | — | 81% |
| Confabulations | — | 22.6% |
| GPQA (HELM) | — | 65% |
Multimodal Not comparable
DeepSeek-V3.2-Exp: —, Llama 4 Maverick: 31.6 (#105)
| Benchmark | DeepSeek-V3.2-Exp | Llama 4 Maverick |
|---|---|---|
| LMArena Vision | — | 1142 |
| GeoBench | — | 52% |
| SpatialViz-Bench | — | 31.8% |
Multilingual DeepSeek-V3.2-Exp leads
DeepSeek-V3.2-Exp: 52.2 (#90), Llama 4 Maverick: 42.2 (#195)
| Benchmark | DeepSeek-V3.2-Exp | Llama 4 Maverick |
|---|---|---|
| LMArena Non-English | 1409 | 1269 |
| LMArena Chinese | 1461 | 1277 |
| LMArena French | 1433 | 1259 |
| LMArena German | 1440 | 1291 |
| LMArena Japanese | 1374 | 1207 |
| LMArena Korean | 1371 | 1203 |
| LMArena Russian | 1424 | 1286 |
| LMArena Spanish | 1440 | 1293 |
Instruction Following DeepSeek-V3.2-Exp leads
DeepSeek-V3.2-Exp: 74.5 (#93), Llama 4 Maverick: 71.7 (#146)
| Benchmark | DeepSeek-V3.2-Exp | Llama 4 Maverick |
|---|---|---|
| LMArena Instruction Following | 1413 | 1267 |
| IFEval | — | 90.8% |
Long Context DeepSeek-V3.2-Exp leads
DeepSeek-V3.2-Exp: 47.6 (#16), Llama 4 Maverick: 31.4 (#279)
| Benchmark | DeepSeek-V3.2-Exp | Llama 4 Maverick |
|---|---|---|
| Fiction.LiveBench | 83.3% | 46.2% |
| LMArena Longer Query | 1428 | 1280 |
| CL-bench | 13.2% | — |
| CL-bench Life | 9.5% | — |
Writing & Preference DeepSeek-V3.2-Exp leads
DeepSeek-V3.2-Exp: 62.4 (#77), Llama 4 Maverick: 38.8 (#252)
| Benchmark | DeepSeek-V3.2-Exp | Llama 4 Maverick |
|---|---|---|
| LMArena Text | 1425 | 1287 |
| LMArena Creative Writing | 1403 | 1267 |
| EQ-Bench Creative Writing | 1515 | 860 |
| LMArena Multi-Turn | 1427 | 1289 |
| Short-Story Creative Writing | — | 62% |
| WildBench | — | 80% |
Frequently asked questions
Is DeepSeek-V3.2-Exp better than Llama 4 Maverick?
DeepSeek-V3.2-Exp is the stronger model overall, scoring 44.3 to 30.9 on the Noometry Index.
Which is cheaper, DeepSeek-V3.2-Exp or Llama 4 Maverick?
DeepSeek-V3.2-Exp is cheaper. It lists at $0.26 per million input tokens and $0.38 per million output tokens; Llama 4 Maverick lists at $0.19 and $0.65.
Is DeepSeek-V3.2-Exp or Llama 4 Maverick better for coding?
DeepSeek-V3.2-Exp scores higher on coding benchmarks: 46.5 versus 26.6 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 4 Maverick share?
36 benchmarks have published results for both models. DeepSeek-V3.2-Exp has 49 scored results on Noometry and Llama 4 Maverick has 54.