Model comparison
GPT-4.1 vs Llama 3.1-70B
GPT-4.1 is the stronger model overall, scoring 35.9 to 29.6 on the Noometry Index. Llama 3.1-70B costs 8.8× less per token, which makes it the better buy when GPT-4.1's lead doesn't matter for your workload.
Last verified . 30 shared benchmarks.
MetaSummary
- They share 30 benchmarks with published results for both. GPT-4.1 scores higher in 8 categories and Llama 3.1-70B in 1 category; 9 gaps are clear of the uncertainty.
- The widest gap is in writing & preference, where GPT-4.1 leads 57.6 to 35.4.
- The biggest single-benchmark swing is MATH Level 5: 83% for GPT-4.1 and 36.7% for Llama 3.1-70B.
- Llama 3.1-70B is cheaper at $0.40 / $0.40 per million input/output tokens, against $2 / $8 for GPT-4.1.
- GPT-4.1 accepts more context: 1.05M tokens versus 128K.
- Llama 3.1-70B has downloadable open weights; the other is API-only.
Side by side
| GPT-4.1 | Llama 3.1-70B | |
|---|---|---|
| Provider | OpenAI | Meta |
| Noometry Index | 35.9 | 29.6 |
| Released | 2025-04-14 | 2024-07-23 |
| Weights | Proprietary | Open |
| Context window | 1.05M | 128K |
| Max output | 33K | 4K |
| Input $ / M tokens | $2 | $0.40 |
| Output $ / M tokens | $8 | $0.40 |
| Results tracked | 52 | 35 |
Sponsored placements are available on pages like this one. Advertise on Noometry
Category by category
Coding GPT-4.1 leads
GPT-4.1: 34.4 (#238), Llama 3.1-70B: 30.3 (#296)
| Benchmark | GPT-4.1 | Llama 3.1-70B |
|---|---|---|
| WeirdML | 39% | 9% |
| LMArena Coding | 1391 | 1260 |
| SWE-bench Verified | 48.5% | — |
| SWE-bench Verified (bash only) | 39.6% | — |
| Aider Polyglot | 52.4% | — |
| BigCodeBench Instruct | — | 46.1% |
| BigCodeBench Complete | — | 54.8% |
| CadEval | 42% | — |
| ALE-Bench | 558.1 | — |
Agentic & Tool Use GPT-4.1 leads
GPT-4.1: 34.7 (#43), Llama 3.1-70B: 25.1 (#112)
| Benchmark | GPT-4.1 | Llama 3.1-70B |
|---|---|---|
| Berkeley Function Calling Leaderboard | 54% | — |
| TheAgentCompany | — | 6.9% |
| BALROG | — | 27.9% |
Reasoning Llama 3.1-70B leads
GPT-4.1: 11.7 (#339), Llama 3.1-70B: 21.6 (#220)
| Benchmark | GPT-4.1 | Llama 3.1-70B |
|---|---|---|
| LMArena Hard Prompts | 1384 | 1241 |
| DTBench | 68.3% | 60% |
| LMCA | 25.6% | 14.8% |
| Epoch Capabilities Index | 136.78 | 125.92 |
| ARC-AGI-2 | 0.4% | — |
| SimpleBench | 27% | — |
| Kagi LLM Benchmark | 52.3% | — |
| ARC-AGI-1 | 5.5% | — |
| Chess Puzzles | 6% | — |
| EnigmaEval | 2.2% | — |
| ForecastBench | 61.5 | — |
Math GPT-4.1 leads
GPT-4.1: 22.3 (#280), Llama 3.1-70B: 13.5 (#304)
| Benchmark | GPT-4.1 | Llama 3.1-70B |
|---|---|---|
| OTIS Mock AIME 2024-2025 | 38.3% | 3.6% |
| Omni-MATH | 47.1% | 21% |
| LMArena Math | 1370 | 1252 |
| MATH Level 5 | 83% | 36.7% |
| FrontierMath (Tiers 1-3) | 6% | — |
| FrontierMath (Feb 2025 set) | 5.5% | — |
| FrontierMath Tier 4 (v1) | 0% | — |
Knowledge GPT-4.1 leads
GPT-4.1: 37.1 (#160), Llama 3.1-70B: 24.2 (#269)
| Benchmark | GPT-4.1 | Llama 3.1-70B |
|---|---|---|
| GPQA Diamond | 66.9% | 44.2% |
| MMLU-Pro | 81.1% | 65.3% |
| GPQA (HELM) | 65.9% | 42.6% |
| LMArena Expert | 1364 | 1209 |
| Humanity's Last Exam | 5.4% | — |
| SimpleQA Verified | 31.1% | — |
| Vectara Hallucination Rate | 5.6% | — |
| MMLU | — | 80.1% |
Multimodal Not comparable
GPT-4.1: 38.2 (#67), Llama 3.1-70B: —
| Benchmark | GPT-4.1 | Llama 3.1-70B |
|---|---|---|
| LMArena Vision | 1211 | — |
| GeoBench | 72% | — |
Multilingual GPT-4.1 leads
GPT-4.1: 49.4 (#133), Llama 3.1-70B: 38.8 (#225)
| Benchmark | GPT-4.1 | Llama 3.1-70B |
|---|---|---|
| LMArena Non-English | 1370 | 1219 |
| LMArena Chinese | 1382 | 1215 |
| LMArena French | 1382 | 1261 |
| LMArena German | 1381 | 1222 |
| LMArena Japanese | 1319 | 1132 |
| LMArena Korean | 1339 | 1140 |
| LMArena Russian | 1377 | 1234 |
| LMArena Spanish | 1376 | 1253 |
Instruction Following GPT-4.1 leads
GPT-4.1: 71.3 (#153), Llama 3.1-70B: 65.3 (#223)
| Benchmark | GPT-4.1 | Llama 3.1-70B |
|---|---|---|
| IFEval | 83.8% | 82.1% |
| LMArena Instruction Following | 1367 | 1231 |
Long Context GPT-4.1 leads
GPT-4.1: 40.0 (#163), Llama 3.1-70B: 37.6 (#214)
| Benchmark | GPT-4.1 | Llama 3.1-70B |
|---|---|---|
| LMArena Longer Query | 1385 | 1241 |
| Fiction.LiveBench | 63.9% | — |
Writing & Preference GPT-4.1 leads
GPT-4.1: 57.6 (#125), Llama 3.1-70B: 35.4 (#267)
| Benchmark | GPT-4.1 | Llama 3.1-70B |
|---|---|---|
| LMArena Text | 1383 | 1261 |
| LMArena Creative Writing | 1363 | 1232 |
| EQ-Bench Creative Writing | 1420 | 784 |
| WildBench | 85.4% | 75.8% |
| LMArena Multi-Turn | 1398 | 1256 |
Frequently asked questions
Is GPT-4.1 better than Llama 3.1-70B?
GPT-4.1 is the stronger model overall, scoring 35.9 to 29.6 on the Noometry Index. Llama 3.1-70B costs 8.8× less per token, which makes it the better buy when GPT-4.1's lead doesn't matter for your workload.
Which is cheaper, GPT-4.1 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-4.1 lists at $2 and $8.
Is GPT-4.1 or Llama 3.1-70B better for coding?
GPT-4.1 scores higher on coding benchmarks: 34.4 versus 30.3 in the Noometry coding category.
Which has the bigger context window?
GPT-4.1 does, with 1.05M tokens against 128K.
How many benchmarks do GPT-4.1 and Llama 3.1-70B share?
30 benchmarks have published results for both models. GPT-4.1 has 52 scored results on Noometry and Llama 3.1-70B has 35.