Model comparison
GPT-4 Turbo vs Llama 3-8B
GPT-4 Turbo is the stronger model overall, scoring 30.3 to 25.3 on the Noometry Index. Llama 3-8B costs 107× less per token, which makes it the better buy when GPT-4 Turbo's lead doesn't matter for your workload.
Last verified . 29 shared benchmarks.
MetaSummary
- They share 29 benchmarks with published results for both. GPT-4 Turbo scores higher in 8 categories and Llama 3-8B in 0 categories; 7 gaps are clear of the uncertainty.
- The widest gap is in knowledge, where GPT-4 Turbo leads 24.3 to 7.8.
- The biggest single-benchmark swing is MATH Level 5: 46.7% for GPT-4 Turbo and 6.1% for Llama 3-8B.
- Llama 3-8B is cheaper at $0.14 / $0.14 per million input/output tokens, against $10 / $30 for GPT-4 Turbo.
- GPT-4 Turbo accepts more context: 128K tokens versus 8K.
- Llama 3-8B has downloadable open weights; the other is API-only.
Side by side
| GPT-4 Turbo | Llama 3-8B | |
|---|---|---|
| Provider | OpenAI | Meta |
| Noometry Index | 30.3 | 25.3 |
| Released | 2023-11-06 | 2024-04-18 |
| Weights | Proprietary | Open |
| Context window | 128K | 8K |
| Max output | 4K | 8K |
| Input $ / M tokens | $10 | $0.14 |
| Output $ / M tokens | $30 | $0.14 |
| Results tracked | 36 | 34 |
Sponsored placements are available on pages like this one. Advertise on Noometry
Category by category
Coding GPT-4 Turbo leads
GPT-4 Turbo: 33.4 (#247), Llama 3-8B: 30.5 (#285)
| Benchmark | GPT-4 Turbo | Llama 3-8B |
|---|---|---|
| BigCodeBench Instruct | 48.2% | 31.9% |
| LMArena Coding | 1268 | 1152 |
| BigCodeBench Complete | 58.2% | 36.9% |
| HumanEval+ | 86.6% | 56.7% |
| MBPP+ | 73.3% | 54.8% |
| WeirdML | 18% | — |
Agentic & Tool Use Not comparable
GPT-4 Turbo: —, Llama 3-8B: —
| Benchmark | GPT-4 Turbo | Llama 3-8B |
|---|---|---|
| METR Time Horizons | 36.7% | — |
Reasoning GPT-4 Turbo leads
GPT-4 Turbo: 14.9 (#311), Llama 3-8B: 13.8 (#321)
| Benchmark | GPT-4 Turbo | Llama 3-8B |
|---|---|---|
| Chess Puzzles | 6% | 0% |
| LMArena Hard Prompts | 1251 | 1133 |
| DTBench | 61.6% | 43.9% |
| Epoch Capabilities Index | 127.25 | 116.33 |
| ForecastBench | 59.4 | 58.6 |
| SimpleBench | 25.1% | — |
| LMCA | 9.8% | — |
| Adversarial NLI | — | 57.3% |
| WinoGrande | — | 75.7% |
Math Too close to call
GPT-4 Turbo: 8.6 (#316), Llama 3-8B: 8.3 (#317)
| Benchmark | GPT-4 Turbo | Llama 3-8B |
|---|---|---|
| OTIS Mock AIME 2024-2025 | 6.7% | 1.9% |
| LMArena Math | 1272 | 1151 |
| MATH Level 5 | 46.7% | 6.1% |
| FrontierMath (Tiers 1-3) | 0.7% | — |
Knowledge GPT-4 Turbo leads
GPT-4 Turbo: 24.3 (#263), Llama 3-8B: 7.8 (#303)
| Benchmark | GPT-4 Turbo | Llama 3-8B |
|---|---|---|
| GPQA Diamond | 46.6% | 26.1% |
| LMArena Expert | 1223 | 1113 |
| MMLU | 81.3% | 68.8% |
| Confabulations | 28.4% | — |
| ARC (AI2) Challenge | — | 82.8% |
| OpenBookQA | — | 82.6% |
| TriviaQA | — | 67.7% |
Multimodal Not comparable
GPT-4 Turbo: 30.0 (#104), Llama 3-8B: —
| Benchmark | GPT-4 Turbo | Llama 3-8B |
|---|---|---|
| LMArena Vision | 1090 | — |
Multilingual GPT-4 Turbo leads
GPT-4 Turbo: 40.7 (#211), Llama 3-8B: 31.0 (#256)
| Benchmark | GPT-4 Turbo | Llama 3-8B |
|---|---|---|
| LMArena Non-English | 1245 | 1098 |
| LMArena Chinese | 1242 | 1076 |
| LMArena French | 1275 | 1158 |
| LMArena German | 1259 | 1104 |
| LMArena Japanese | 1194 | 967 |
| LMArena Korean | 1187 | 1003 |
| LMArena Russian | 1259 | 1109 |
| LMArena Spanish | 1260 | 1172 |
Instruction Following GPT-4 Turbo leads
GPT-4 Turbo: 66.0 (#211), Llama 3-8B: 58.7 (#255)
| Benchmark | GPT-4 Turbo | Llama 3-8B |
|---|---|---|
| LMArena Instruction Following | 1250 | 1127 |
Long Context GPT-4 Turbo leads
GPT-4 Turbo: 38.0 (#200), Llama 3-8B: 34.2 (#246)
| Benchmark | GPT-4 Turbo | Llama 3-8B |
|---|---|---|
| LMArena Longer Query | 1254 | 1128 |
Writing & Preference GPT-4 Turbo leads
GPT-4 Turbo: 47.9 (#201), Llama 3-8B: 37.7 (#251)
| Benchmark | GPT-4 Turbo | Llama 3-8B |
|---|---|---|
| LMArena Text | 1272 | 1166 |
| LMArena Creative Writing | 1269 | 1150 |
| LMArena Multi-Turn | 1267 | 1152 |
Frequently asked questions
Is GPT-4 Turbo better than Llama 3-8B?
GPT-4 Turbo is the stronger model overall, scoring 30.3 to 25.3 on the Noometry Index. Llama 3-8B costs 107× less per token, which makes it the better buy when GPT-4 Turbo's lead doesn't matter for your workload.
Which is cheaper, GPT-4 Turbo or Llama 3-8B?
Llama 3-8B is cheaper. It lists at $0.14 per million input tokens and $0.14 per million output tokens; GPT-4 Turbo lists at $10 and $30.
Is GPT-4 Turbo or Llama 3-8B better for coding?
GPT-4 Turbo scores higher on coding benchmarks: 33.4 versus 30.5 in the Noometry coding category.
Which has the bigger context window?
GPT-4 Turbo does, with 128K tokens against 8K.
How many benchmarks do GPT-4 Turbo and Llama 3-8B share?
29 benchmarks have published results for both models. GPT-4 Turbo has 36 scored results on Noometry and Llama 3-8B has 34.