Model comparison
GPT-5.2 vs Llama 3.1-70B
GPT-5.2 is the stronger model overall, scoring 54.1 to 29.4 on the Noometry Index. Llama 3.1-70B costs 12× less per token, which makes it the better buy when GPT-5.2's lead doesn't matter for your workload.
Last verified . 24 shared benchmarks.
MetaSummary
- They share 24 benchmarks with published results for both. GPT-5.2 scores higher in 9 categories and Llama 3.1-70B in 0 categories; 9 gaps are clear of the uncertainty.
- The widest gap is in math, where GPT-5.2 leads 59.8 to 12.7.
- The biggest single-benchmark swing is OTIS Mock AIME 2024-2025: 96.1% for GPT-5.2 and 3.6% for Llama 3.1-70B.
- Llama 3.1-70B is cheaper at $0.40 / $0.40 per million input/output tokens, against $1.75 / $14 for GPT-5.2.
- GPT-5.2 accepts more context: 400K tokens versus 128K.
- Llama 3.1-70B has downloadable open weights; the other is API-only.
Side by side
| GPT-5.2 | Llama 3.1-70B | |
|---|---|---|
| Provider | OpenAI | Meta |
| Noometry Index | 54.1 | 29.4 |
| Released | 2025-12-11 | 2024-07-23 |
| Weights | Proprietary | Open |
| Context window | 400K | 128K |
| Max output | 128K | 4K |
| Input $ / M tokens | $1.75 | $0.40 |
| Output $ / M tokens | $14 | $0.40 |
| Results tracked | 67 | 35 |
Sponsored placements are available on pages like this one. Advertise on Noometry
Category by category
Coding GPT-5.2 leads
GPT-5.2: 52.2 (#32), Llama 3.1-70B: 29.8 (#292)
| Benchmark | GPT-5.2 | Llama 3.1-70B |
|---|---|---|
| WeirdML | 72.2% | 9% |
| LMArena Coding | 1447 | 1260 |
| SWE-bench Verified | 73.8% | — |
| SWE-bench Verified (bash only) | 72.8% | — |
| LMArena WebDev | 1415 | — |
| SWE-bench Multilingual | 66.7% | — |
| GSO | 27.4% | — |
| BigCodeBench Instruct | — | 46.1% |
| BigCodeBench Complete | — | 54.8% |
| ALE-Bench | 1,294 | — |
| AlgoTune | 2.05 | — |
Agentic & Tool Use GPT-5.2 leads
GPT-5.2: 40.0 (#24), Llama 3.1-70B: 24.9 (#104)
| Benchmark | GPT-5.2 | Llama 3.1-70B |
|---|---|---|
| Terminal-Bench | 64.9% | — |
| Berkeley Function Calling Leaderboard | 55.9% | — |
| GDPval | 49.7% | — |
| Remote Labor Index | 2.5% | — |
| TheAgentCompany | — | 6.9% |
| τ²-bench Airline | 83% | — |
| τ²-bench Banking | 32.2% | — |
| τ²-bench Retail | 81.6% | — |
| τ²-bench Telecom | 89.7% | — |
| DeepResearch Bench | 41.1% | — |
| BALROG | — | 27.9% |
| LMArena Search | 1207 | — |
| METR Time Horizons | 75.3% | — |
| Vending-Bench 2 | 3,591 | — |
Reasoning GPT-5.2 leads
GPT-5.2: 50.1 (#33), Llama 3.1-70B: 20.8 (#215)
| Benchmark | GPT-5.2 | Llama 3.1-70B |
|---|---|---|
| LMArena Hard Prompts | 1445 | 1241 |
| DTBench | 90.9% | 60% |
| LMCA | 43.9% | 14.8% |
| Epoch Capabilities Index | 153.45 | 125.91 |
| ARC-AGI-2 | 52.9% | — |
| SimpleBench | 45.8% | — |
| Kagi LLM Benchmark | 73.3% | — |
| NYT Connections (extended) | 83.6% | — |
| ARC-AGI-1 | 86.2% | — |
| Chess Puzzles | 49% | — |
| EnigmaEval | 10.4% | — |
| EBR-Bench | 23% | — |
| Mystery Game Puzzles | 23% | — |
| ForecastBench | 60.1 | — |
Math GPT-5.2 leads
GPT-5.2: 59.8 (#32), Llama 3.1-70B: 12.7 (#298)
| Benchmark | GPT-5.2 | Llama 3.1-70B |
|---|---|---|
| OTIS Mock AIME 2024-2025 | 96.1% | 3.6% |
| LMArena Math | 1439 | 1252 |
| FrontierMath (Tiers 1-3) | 67.4% | — |
| FrontierMath Tier 4 | 31.7% | — |
| MathArena Final-Answer Competitions | 72% | — |
| ProofBench | 15% | — |
| Omni-MATH | — | 21% |
| MATH Level 5 | — | 36.7% |
| FrontierMath (Feb 2025 set) | 40.7% | — |
| FrontierMath Tier 4 (v1) | 18.8% | — |
Knowledge GPT-5.2 leads
GPT-5.2: 59.2 (#29), Llama 3.1-70B: 24.1 (#264)
| Benchmark | GPT-5.2 | Llama 3.1-70B |
|---|---|---|
| GPQA Diamond | 91.4% | 44.2% |
| LMArena Expert | 1445 | 1209 |
| Humanity's Last Exam | 27.8% | — |
| SimpleQA Verified | 37.1% | — |
| MMLU-Pro | — | 65.3% |
| Vectara Hallucination Rate | 8.4% | — |
| GPQA (HELM) | — | 42.6% |
| MMLU | — | 80.1% |
Multimodal Not comparable
GPT-5.2: 50.9 (#7), Llama 3.1-70B: —
| Benchmark | GPT-5.2 | Llama 3.1-70B |
|---|---|---|
| LMArena Vision | 1268 | — |
| VPCT | 84% | — |
| Furniture Assembly | 38.3% | — |
| LMArena Document | 1405 | — |
Multilingual GPT-5.2 leads
GPT-5.2: 53.6 (#63), Llama 3.1-70B: 38.9 (#220)
| Benchmark | GPT-5.2 | Llama 3.1-70B |
|---|---|---|
| LMArena Non-English | 1425 | 1219 |
| LMArena Chinese | 1460 | 1215 |
| LMArena French | 1456 | 1260 |
| LMArena German | 1448 | 1222 |
| LMArena Japanese | 1421 | 1132 |
| LMArena Korean | 1390 | 1140 |
| LMArena Russian | 1440 | 1234 |
| LMArena Spanish | 1433 | 1253 |
Instruction Following GPT-5.2 leads
GPT-5.2: 74.9 (#83), Llama 3.1-70B: 65.5 (#218)
| Benchmark | GPT-5.2 | Llama 3.1-70B |
|---|---|---|
| LMArena Instruction Following | 1417 | 1231 |
| IFEval | — | 82.1% |
Long Context GPT-5.2 leads
GPT-5.2: 44.0 (#74), Llama 3.1-70B: 37.6 (#208)
| Benchmark | GPT-5.2 | Llama 3.1-70B |
|---|---|---|
| LMArena Longer Query | 1428 | 1241 |
| CL-bench | 18.2% | — |
Writing & Preference GPT-5.2 leads
GPT-5.2: 67.0 (#30), Llama 3.1-70B: 35.6 (#262)
| Benchmark | GPT-5.2 | Llama 3.1-70B |
|---|---|---|
| LMArena Text | 1439 | 1261 |
| LMArena Creative Writing | 1402 | 1232 |
| EQ-Bench Creative Writing | 1703 | 784 |
| LMArena Multi-Turn | 1458 | 1256 |
| WildBench | — | 75.8% |
Frequently asked questions
Is GPT-5.2 better than Llama 3.1-70B?
GPT-5.2 is the stronger model overall, scoring 54.1 to 29.4 on the Noometry Index. Llama 3.1-70B costs 12× 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 3.1-70B?
Llama 3.1-70B is cheaper. It lists at $0.40 per million input tokens and $0.40 per million output tokens; GPT-5.2 lists at $1.75 and $14.
Is GPT-5.2 or Llama 3.1-70B better for coding?
GPT-5.2 scores higher on coding benchmarks: 52.2 versus 29.8 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 3.1-70B share?
24 benchmarks have published results for both models. GPT-5.2 has 67 scored results on Noometry and Llama 3.1-70B has 35.