Noometry

Model comparison

GPT-4 vs Llama 3.1-405B

Llama 3.1-405B is the stronger model overall, scoring 30.7 to 29.1 on the Noometry Index.

Last verified . 30 shared benchmarks.

GPT-4 OpenAI

29.1

Rank #316 Confirmed

Llama 3.1-405B Meta

30.7

Rank #288 Confirmed

Summary

  • They share 30 benchmarks with published results for both. GPT-4 scores higher in 1 category and Llama 3.1-405B in 7 categories; 4 gaps are clear of the uncertainty.
  • The widest gap is in knowledge, where Llama 3.1-405B leads 30.4 to 18.4.
  • The biggest single-benchmark swing is MATH Level 5: 23% for GPT-4 and 49.8% for Llama 3.1-405B.
  • Llama 3.1-405B has downloadable open weights; the other is API-only.

Side by side

GPT-4Llama 3.1-405B
ProviderOpenAIMeta
Noometry Index29.130.7
Released2023-03-142024-07-23
WeightsProprietaryOpen
Context window8K—
Max output8K—
Input $ / M tokens$30—
Output $ / M tokens$60—
Results tracked3842

Sponsored placements are available on pages like this one. Advertise on Noometry

Category by category

Coding Llama 3.1-405B leads

GPT-4: 31.6 (#283), Llama 3.1-405B: 33.1 (#262)

BenchmarkGPT-4Llama 3.1-405B
WeirdML12.4%21.4%
LMArena Coding12541291
BigCodeBench Instruct46%—
BigCodeBench Complete57.2%—
HumanEval+79.3%—

Agentic & Tool Use Not comparable

GPT-4: —, Llama 3.1-405B: 21.0 (#140)

BenchmarkGPT-4Llama 3.1-405B
TheAgentCompany—7.4%
Cybench—7.5%
METR Time Horizons36.1%—

Reasoning Too close to call

GPT-4: 17.8 (#289), Llama 3.1-405B: 16.8 (#300)

BenchmarkGPT-4Llama 3.1-405B
LMArena Hard Prompts12411269
DTBench62.7%61.4%
BIG-Bench Hard75.1%82.9%
Epoch Capabilities Index125.89128.75
ForecastBench57.859.9
HellaSwag95.3%89.2%
WinoGrande87.5%89.2%
SimpleBench—23%
Kagi LLM Benchmark—45%
Chess Puzzles4%—
Mystery Game Puzzles12%—
LMCA17.1%—
PIQA—85.9%

Math Llama 3.1-405B leads

GPT-4: 10.8 (#309), Llama 3.1-405B: 18.4 (#290)

BenchmarkGPT-4Llama 3.1-405B
OTIS Mock AIME 2024-20251.1%9.7%
LMArena Math12691281
MATH Level 523%49.8%
Omni-MATH—24.9%
GSM8K92%—

Knowledge Llama 3.1-405B leads

GPT-4: 18.4 (#282), Llama 3.1-405B: 30.4 (#227)

BenchmarkGPT-4Llama 3.1-405B
GPQA Diamond35.7%50.9%
LMArena Expert12111243
MMLU86.4%84.5%
TriviaQA84.8%82.7%
MMLU-Pro—72.3%
Confabulations—17.6%
GPQA (HELM)—52.2%
ARC (AI2) Challenge—95.3%

Multilingual Too close to call

GPT-4: 40.6 (#215), Llama 3.1-405B: 40.7 (#214)

BenchmarkGPT-4Llama 3.1-405B
LMArena Non-English12461248
LMArena Chinese12421242
LMArena French12831279
LMArena German12511252
LMArena Japanese12091208
LMArena Korean11841184
LMArena Russian12511265
LMArena Spanish12611260

Instruction Following Too close to call

GPT-4: 65.3 (#222), Llama 3.1-405B: 65.9 (#214)

BenchmarkGPT-4Llama 3.1-405B
LMArena Instruction Following12411259
IFEval—81.1%

Long Context Too close to call

GPT-4: 37.7 (#212), Llama 3.1-405B: 38.4 (#197)

BenchmarkGPT-4Llama 3.1-405B
LMArena Longer Query12441266

Writing & Preference Llama 3.1-405B leads

GPT-4: 34.9 (#268), Llama 3.1-405B: 38.9 (#251)

BenchmarkGPT-4Llama 3.1-405B
LMArena Text12631284
LMArena Creative Writing12441262
EQ-Bench Creative Writing752870
LMArena Multi-Turn12571297
WildBench—78.3%

Frequently asked questions

Is GPT-4 better than Llama 3.1-405B?

Llama 3.1-405B is the stronger model overall, scoring 30.7 to 29.1 on the Noometry Index.

Is GPT-4 or Llama 3.1-405B better for coding?

Llama 3.1-405B scores higher on coding benchmarks: 33.1 versus 31.6 in the Noometry coding category.

How many benchmarks do GPT-4 and Llama 3.1-405B share?

30 benchmarks have published results for both models. GPT-4 has 38 scored results on Noometry and Llama 3.1-405B has 42.

Related comparisons

Go deeper