Model comparison
Llama 3.1-70B vs Llama 3-8B
Llama 3.1-70B is the stronger model overall, scoring 29.4 to 25.3 on the Noometry Index. Llama 3-8B costs 2.9× less per token, which makes it the better buy when Llama 3.1-70B's lead doesn't matter for your workload.
Last verified . 25 shared benchmarks.
MetaSummary
- They share 25 benchmarks with published results for both. Llama 3.1-70B scores higher in 6 categories and Llama 3-8B in 2 categories; 7 gaps are clear of the uncertainty.
- The widest gap is in knowledge, where Llama 3.1-70B leads 24.1 to 7.8.
- The biggest single-benchmark swing is MATH Level 5: 36.7% for Llama 3.1-70B and 6.1% for Llama 3-8B.
- Llama 3-8B is cheaper at $0.14 / $0.14 per million input/output tokens, against $0.40 / $0.40 for Llama 3.1-70B.
- Llama 3.1-70B accepts more context: 128K tokens versus 8K.
Side by side
| Llama 3.1-70B | Llama 3-8B | |
|---|---|---|
| Provider | Meta | Meta |
| Noometry Index | 29.4 | 25.3 |
| Released | 2024-07-23 | 2024-04-18 |
| Weights | Open | Open |
| Context window | 128K | 8K |
| Max output | 4K | 8K |
| Input $ / M tokens | $0.40 | $0.14 |
| Output $ / M tokens | $0.40 | $0.14 |
| Results tracked | 35 | 34 |
Sponsored placements are available on pages like this one. Advertise on Noometry
Category by category
Coding Too close to call
Llama 3.1-70B: 29.8 (#292), Llama 3-8B: 30.5 (#285)
| Benchmark | Llama 3.1-70B | Llama 3-8B |
|---|---|---|
| BigCodeBench Instruct | 46.1% | 31.9% |
| LMArena Coding | 1260 | 1152 |
| BigCodeBench Complete | 54.8% | 36.9% |
| WeirdML | 9% | — |
| HumanEval+ | — | 56.7% |
| MBPP+ | — | 54.8% |
Agentic & Tool Use Not comparable
Llama 3.1-70B: 24.9 (#104), Llama 3-8B: —
| Benchmark | Llama 3.1-70B | Llama 3-8B |
|---|---|---|
| TheAgentCompany | 6.9% | — |
| BALROG | 27.9% | — |
Reasoning Llama 3.1-70B leads
Llama 3.1-70B: 20.8 (#215), Llama 3-8B: 13.8 (#321)
| Benchmark | Llama 3.1-70B | Llama 3-8B |
|---|---|---|
| LMArena Hard Prompts | 1241 | 1133 |
| DTBench | 60% | 43.9% |
| Epoch Capabilities Index | 125.91 | 116.33 |
| Chess Puzzles | — | 0% |
| LMCA | 14.8% | — |
| Adversarial NLI | — | 57.3% |
| ForecastBench | — | 58.6 |
| WinoGrande | — | 75.7% |
Math Llama 3.1-70B leads
Llama 3.1-70B: 12.7 (#298), Llama 3-8B: 8.3 (#317)
| Benchmark | Llama 3.1-70B | Llama 3-8B |
|---|---|---|
| OTIS Mock AIME 2024-2025 | 3.6% | 1.9% |
| LMArena Math | 1252 | 1151 |
| MATH Level 5 | 36.7% | 6.1% |
| Omni-MATH | 21% | — |
Knowledge Llama 3.1-70B leads
Llama 3.1-70B: 24.1 (#264), Llama 3-8B: 7.8 (#303)
| Benchmark | Llama 3.1-70B | Llama 3-8B |
|---|---|---|
| GPQA Diamond | 44.2% | 26.1% |
| LMArena Expert | 1209 | 1113 |
| MMLU | 80.1% | 68.8% |
| MMLU-Pro | 65.3% | — |
| GPQA (HELM) | 42.6% | — |
| ARC (AI2) Challenge | — | 82.8% |
| OpenBookQA | — | 82.6% |
| TriviaQA | — | 67.7% |
Multilingual Llama 3.1-70B leads
Llama 3.1-70B: 38.9 (#220), Llama 3-8B: 31.0 (#256)
| Benchmark | Llama 3.1-70B | Llama 3-8B |
|---|---|---|
| LMArena Non-English | 1219 | 1098 |
| LMArena Chinese | 1215 | 1076 |
| LMArena French | 1260 | 1158 |
| LMArena German | 1222 | 1104 |
| LMArena Japanese | 1132 | 967 |
| LMArena Korean | 1140 | 1003 |
| LMArena Russian | 1234 | 1109 |
| LMArena Spanish | 1253 | 1172 |
Instruction Following Llama 3.1-70B leads
Llama 3.1-70B: 65.5 (#218), Llama 3-8B: 58.7 (#255)
| Benchmark | Llama 3.1-70B | Llama 3-8B |
|---|---|---|
| LMArena Instruction Following | 1231 | 1127 |
| IFEval | 82.1% | — |
Long Context Llama 3.1-70B leads
Llama 3.1-70B: 37.6 (#208), Llama 3-8B: 34.2 (#246)
| Benchmark | Llama 3.1-70B | Llama 3-8B |
|---|---|---|
| LMArena Longer Query | 1241 | 1128 |
Writing & Preference Llama 3-8B leads
Llama 3.1-70B: 35.6 (#262), Llama 3-8B: 37.7 (#251)
| Benchmark | Llama 3.1-70B | Llama 3-8B |
|---|---|---|
| LMArena Text | 1261 | 1166 |
| LMArena Creative Writing | 1232 | 1150 |
| LMArena Multi-Turn | 1256 | 1152 |
| EQ-Bench Creative Writing | 784 | — |
| WildBench | 75.8% | — |
Frequently asked questions
Is Llama 3.1-70B better than Llama 3-8B?
Llama 3.1-70B is the stronger model overall, scoring 29.4 to 25.3 on the Noometry Index. Llama 3-8B costs 2.9× less per token, which makes it the better buy when Llama 3.1-70B's lead doesn't matter for your workload.
Which is cheaper, Llama 3.1-70B 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; Llama 3.1-70B lists at $0.40 and $0.40.
Is Llama 3.1-70B or Llama 3-8B better for coding?
They score almost the same on coding (29.8 vs 30.5); test both on your own repository before choosing.
Which has the bigger context window?
Llama 3.1-70B does, with 128K tokens against 8K.
How many benchmarks do Llama 3.1-70B and Llama 3-8B share?
25 benchmarks have published results for both models. Llama 3.1-70B has 35 scored results on Noometry and Llama 3-8B has 34.