Model comparison

Llama 3.1-405B vs StarCoder 2 15B

Llama 3.1-405B has enough public results to be ranked (#288); StarCoder 2 15B does not yet, so treat this comparison as directional.

Last verified . 4 shared benchmarks.

Llama 3.1-405B Meta

30.7

Rank #288 Confirmed

StarCoder 2 15B NVIDIA

38.1

Unranked Sparse

Summary

  • They share 4 benchmarks with published results for both. Llama 3.1-405B scores higher in 0 categories and StarCoder 2 15B in 1 category; one gap is clear of the uncertainty.
  • The widest gap is in coding, where StarCoder 2 15B leads 36.6 to 33.1.

Side by side

Llama 3.1-405B and StarCoder 2 15B specifications
Llama 3.1-405BStarCoder 2 15B
ProviderMetaNVIDIA
Noometry Index30.738.1
Released2024-07-232024-02-20
WeightsOpenOpen
Context window——
Max output——
Input $ / M tokens——
Output $ / M tokens——
Results tracked427

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

Category by category

Coding StarCoder 2 15B leads

Llama 3.1-405B: 33.1 (#262), StarCoder 2 15B: 36.6

Coding benchmarks
BenchmarkLlama 3.1-405BStarCoder 2 15B
WeirdML21.4%—
LMArena Coding1291—
BigCodeBench Complete—38.4%
HumanEval+—37.8%

Agentic & Tool Use Not comparable

Llama 3.1-405B: 21.0 (#140), StarCoder 2 15B: —

Agentic & Tool Use benchmarks
BenchmarkLlama 3.1-405BStarCoder 2 15B
TheAgentCompany7.4%—
Cybench7.5%—

Reasoning Not comparable

Llama 3.1-405B: 16.8 (#300), StarCoder 2 15B: —

Reasoning benchmarks
BenchmarkLlama 3.1-405BStarCoder 2 15B
Epoch Capabilities Index128.75105
WinoGrande89.2%64.3%
SimpleBench23%—
Kagi LLM Benchmark45%—
LMArena Hard Prompts1269—
DTBench61.4%—
BIG-Bench Hard82.9%—
ForecastBench59.9—
HellaSwag89.2%—
PIQA85.9%—

Math Not comparable

Llama 3.1-405B: 18.4 (#290), StarCoder 2 15B: —

Math benchmarks
BenchmarkLlama 3.1-405BStarCoder 2 15B
OTIS Mock AIME 2024-20259.7%—
Omni-MATH24.9%—
LMArena Math1281—
MATH Level 549.8%—
GSM8K—57.7%

Knowledge Not comparable

Llama 3.1-405B: 30.4 (#227), StarCoder 2 15B: —

Knowledge benchmarks
BenchmarkLlama 3.1-405BStarCoder 2 15B
ARC (AI2) Challenge95.3%47.2%
MMLU84.5%64.1%
GPQA Diamond50.9%—
MMLU-Pro72.3%—
Confabulations17.6%—
GPQA (HELM)52.2%—
LMArena Expert1243—
TriviaQA82.7%—

Multilingual Not comparable

Llama 3.1-405B: 40.7 (#214), StarCoder 2 15B: —

Multilingual benchmarks
BenchmarkLlama 3.1-405BStarCoder 2 15B
LMArena Non-English1248—
LMArena Chinese1242—
LMArena French1279—
LMArena German1252—
LMArena Japanese1208—
LMArena Korean1184—
LMArena Russian1265—
LMArena Spanish1260—

Instruction Following Not comparable

Llama 3.1-405B: 65.9 (#214), StarCoder 2 15B: —

Instruction Following benchmarks
BenchmarkLlama 3.1-405BStarCoder 2 15B
IFEval81.1%—
LMArena Instruction Following1259—

Long Context Not comparable

Llama 3.1-405B: 38.4 (#197), StarCoder 2 15B: —

Long Context benchmarks
BenchmarkLlama 3.1-405BStarCoder 2 15B
LMArena Longer Query1266—

Writing & Preference Not comparable

Llama 3.1-405B: 38.9 (#251), StarCoder 2 15B: —

Writing & Preference benchmarks
BenchmarkLlama 3.1-405BStarCoder 2 15B
LMArena Text1284—
LMArena Creative Writing1262—
EQ-Bench Creative Writing870—
WildBench78.3%—
LMArena Multi-Turn1297—

Frequently asked questions

Is Llama 3.1-405B better than StarCoder 2 15B?

Llama 3.1-405B has enough public results to be ranked (#288); StarCoder 2 15B does not yet, so treat this comparison as directional.

Is Llama 3.1-405B or StarCoder 2 15B better for coding?

StarCoder 2 15B scores higher on coding benchmarks: 36.6 versus 33.1 in the Noometry coding category.

How many benchmarks do Llama 3.1-405B and StarCoder 2 15B share?

4 benchmarks have published results for both models. Llama 3.1-405B has 42 scored results on Noometry and StarCoder 2 15B has 7.

Related comparisons

Go deeper