Model comparison
GPT-5.2 vs Llama 4 Maverick
GPT-5.2 is the stronger model overall, scoring 54.1 to 30.9 on the Noometry Index. Llama 4 Maverick costs 16× less per token, which makes it the better buy when GPT-5.2's lead doesn't matter for your workload.
Last verified . 38 shared benchmarks.
MetaSummary
- They share 38 benchmarks with published results for both. GPT-5.2 scores higher in 10 categories and Llama 4 Maverick in 0 categories; 10 gaps are clear of the uncertainty.
- The widest gap is in reasoning, where GPT-5.2 leads 50.2 to 10.1.
- The biggest single-benchmark swing is ARC-AGI-1: 86.2% for GPT-5.2 and 4.4% for Llama 4 Maverick.
- Llama 4 Maverick is cheaper at $0.19 / $0.65 per million input/output tokens, against $1.75 / $14 for GPT-5.2.
- GPT-5.2 accepts more context: 400K tokens versus 128K.
- Llama 4 Maverick has downloadable open weights; the other is API-only.
Side by side
| GPT-5.2 | Llama 4 Maverick | |
|---|---|---|
| Provider | OpenAI | Meta |
| Noometry Index | 54.1 | 30.9 |
| Released | 2025-12-11 | 2025-04-05 |
| Weights | Proprietary | Open |
| Context window | 400K | 128K |
| Max output | 128K | 4K |
| Input $ / M tokens | $1.75 | $0.19 |
| Output $ / M tokens | $14 | $0.65 |
| Results tracked | 67 | 54 |
Sponsored placements are available on pages like this one. Advertise on Noometry
Category by category
Coding GPT-5.2 leads
GPT-5.2: 51.6 (#37), Llama 4 Maverick: 26.6 (#324)
| Benchmark | GPT-5.2 | Llama 4 Maverick |
|---|---|---|
| SWE-bench Verified (bash only) | 72.8% | 21% |
| WeirdML | 72.2% | 24.5% |
| LMArena Coding | 1447 | 1302 |
| ALE-Bench | 1,294 | 172.97 |
| SWE-bench Verified | 73.8% | — |
| Aider Polyglot | — | 15.6% |
| LMArena WebDev | 1416 | — |
| SWE-bench Multilingual | 66.7% | — |
| SciCode | — | 33.1% |
| GSO | 27.4% | — |
| BigCodeBench Instruct | — | 49.7% |
| BigCodeBench Complete | — | 61.4% |
| AlgoTune | 2.05 | — |
Agentic & Tool Use GPT-5.2 leads
GPT-5.2: 40.2 (#24), Llama 4 Maverick: 28.2 (#91)
| Benchmark | GPT-5.2 | Llama 4 Maverick |
|---|---|---|
| Berkeley Function Calling Leaderboard | 55.9% | 37.3% |
| Terminal-Bench | 64.9% | — |
| GDPval | 49.7% | — |
| Remote Labor Index | 2.5% | — |
| τ²-bench Airline | 83% | — |
| τ²-bench Banking | 32.2% | — |
| τ²-bench Retail | 81.6% | — |
| τ²-bench Telecom | 89.7% | — |
| DeepResearch Bench | 41.1% | — |
| LMArena Search | 1207 | — |
| METR Time Horizons | 75.3% | — |
| Vending-Bench 2 | 3,591 | — |
Reasoning GPT-5.2 leads
GPT-5.2: 50.2 (#35), Llama 4 Maverick: 10.1 (#342)
| Benchmark | GPT-5.2 | Llama 4 Maverick |
|---|---|---|
| ARC-AGI-2 | 52.9% | 0% |
| SimpleBench | 45.8% | 27.7% |
| Kagi LLM Benchmark | 73.3% | 55.9% |
| NYT Connections (extended) | 83.6% | 8% |
| ARC-AGI-1 | 86.2% | 4.4% |
| EnigmaEval | 10.4% | 0.6% |
| LMArena Hard Prompts | 1445 | 1281 |
| DTBench | 90.9% | 61.9% |
| LMCA | 43.9% | 15.9% |
| Epoch Capabilities Index | 153.45 | 132.2 |
| ForecastBench | 60.1 | 57.5 |
| CritPt | — | 0% |
| Chess Puzzles | 49% | — |
| EBR-Bench | 23% | — |
| Mystery Game Puzzles | 23% | — |
Math GPT-5.2 leads
GPT-5.2: 60.0 (#38), Llama 4 Maverick: 26.0 (#262)
| Benchmark | GPT-5.2 | Llama 4 Maverick |
|---|---|---|
| OTIS Mock AIME 2024-2025 | 96.1% | 20.6% |
| LMArena Math | 1440 | 1299 |
| FrontierMath (Feb 2025 set) | 40.7% | 0.7% |
| FrontierMath (Tiers 1-3) | 67.4% | — |
| FrontierMath Tier 4 | 31.7% | — |
| MathArena Final-Answer Competitions | 72% | — |
| ProofBench | 15% | — |
| Omni-MATH | — | 42.2% |
| MATH Level 5 | — | 73% |
| FrontierMath Tier 4 (v1) | 18.8% | — |
Knowledge GPT-5.2 leads
GPT-5.2: 59.3 (#32), Llama 4 Maverick: 33.4 (#204)
| Benchmark | GPT-5.2 | Llama 4 Maverick |
|---|---|---|
| GPQA Diamond | 91.4% | 67% |
| Humanity's Last Exam | 27.8% | 5.7% |
| Vectara Hallucination Rate | 8.4% | 8.2% |
| LMArena Expert | 1445 | 1259 |
| SimpleQA Verified | 37.1% | — |
| MMLU-Pro | — | 81% |
| Confabulations | — | 22.6% |
| GPQA (HELM) | — | 65% |
Multimodal GPT-5.2 leads
GPT-5.2: 51.3 (#7), Llama 4 Maverick: 31.6 (#105)
| Benchmark | GPT-5.2 | Llama 4 Maverick |
|---|---|---|
| LMArena Vision | 1268 | 1142 |
| GeoBench | — | 52% |
| VPCT | 84% | — |
| Furniture Assembly | 38.3% | — |
| LMArena Document | 1405 | — |
| SpatialViz-Bench | — | 31.8% |
Multilingual GPT-5.2 leads
GPT-5.2: 53.4 (#67), Llama 4 Maverick: 42.2 (#195)
| Benchmark | GPT-5.2 | Llama 4 Maverick |
|---|---|---|
| LMArena Non-English | 1425 | 1269 |
| LMArena Chinese | 1460 | 1277 |
| LMArena French | 1455 | 1259 |
| LMArena German | 1448 | 1291 |
| LMArena Japanese | 1420 | 1207 |
| LMArena Korean | 1392 | 1203 |
| LMArena Russian | 1440 | 1286 |
| LMArena Spanish | 1433 | 1293 |
Instruction Following GPT-5.2 leads
GPT-5.2: 74.7 (#89), Llama 4 Maverick: 71.7 (#146)
| Benchmark | GPT-5.2 | Llama 4 Maverick |
|---|---|---|
| LMArena Instruction Following | 1417 | 1267 |
| IFEval | — | 90.8% |
Long Context GPT-5.2 leads
GPT-5.2: 44.0 (#78), Llama 4 Maverick: 31.4 (#279)
| Benchmark | GPT-5.2 | Llama 4 Maverick |
|---|---|---|
| LMArena Longer Query | 1428 | 1280 |
| Fiction.LiveBench | — | 46.2% |
| CL-bench | 18.2% | — |
Writing & Preference GPT-5.2 leads
GPT-5.2: 66.8 (#32), Llama 4 Maverick: 38.8 (#252)
| Benchmark | GPT-5.2 | Llama 4 Maverick |
|---|---|---|
| LMArena Text | 1439 | 1287 |
| LMArena Creative Writing | 1401 | 1267 |
| EQ-Bench Creative Writing | 1703 | 860 |
| LMArena Multi-Turn | 1458 | 1289 |
| Short-Story Creative Writing | — | 62% |
| WildBench | — | 80% |
Frequently asked questions
Is GPT-5.2 better than Llama 4 Maverick?
GPT-5.2 is the stronger model overall, scoring 54.1 to 30.9 on the Noometry Index. Llama 4 Maverick costs 16× less per token, which makes it the better buy when GPT-5.2's lead doesn't matter for your workload.
Which is cheaper, GPT-5.2 or Llama 4 Maverick?
Llama 4 Maverick is cheaper. It lists at $0.19 per million input tokens and $0.65 per million output tokens; GPT-5.2 lists at $1.75 and $14.
Is GPT-5.2 or Llama 4 Maverick better for coding?
GPT-5.2 scores higher on coding benchmarks: 51.6 versus 26.6 in the Noometry coding category.
Which has the bigger context window?
GPT-5.2 does, with 400K tokens against 128K.
How many benchmarks do GPT-5.2 and Llama 4 Maverick share?
38 benchmarks have published results for both models. GPT-5.2 has 67 scored results on Noometry and Llama 4 Maverick has 54.