Model comparison
Llama-3.3-70B-Instruct vs Llama 3-8B
Llama-3.3-70B-Instruct is the stronger model overall, scoring 30.3 to 25.3 on the Noometry Index.
Last verified . 26 shared benchmarks.
MetaSummary
- They share 26 benchmarks with published results for both. Llama-3.3-70B-Instruct scores higher in 6 categories and Llama 3-8B in 2 categories; 6 gaps are clear of the uncertainty.
- The widest gap is in knowledge, where Llama-3.3-70B-Instruct leads 30.5 to 7.8.
- The biggest single-benchmark swing is MATH Level 5: 41.6% for Llama-3.3-70B-Instruct and 6.1% for Llama 3-8B.
- Llama 3-8B is cheaper at $0.14 / $0.14 per million input/output tokens, against $0.10 / $0.32 for Llama-3.3-70B-Instruct.
- Llama-3.3-70B-Instruct accepts more context: 128K tokens versus 8K.
Side by side
| Llama-3.3-70B-Instruct | Llama 3-8B | |
|---|---|---|
| Provider | Meta | Meta |
| Noometry Index | 30.3 | 25.3 |
| Released | 2024-12-06 | 2024-04-18 |
| Weights | Open | Open |
| Context window | 128K | 8K |
| Max output | 4K | 8K |
| Input $ / M tokens | $0.10 | $0.14 |
| Output $ / M tokens | $0.32 | $0.14 |
| Results tracked | 43 | 34 |
Sponsored placements are available on pages like this one. Advertise on Noometry
Category by category
Coding Too close to call
Llama-3.3-70B-Instruct: 30.6 (#284), Llama 3-8B: 30.5 (#285)
| Benchmark | Llama-3.3-70B-Instruct | Llama 3-8B |
|---|---|---|
| BigCodeBench Instruct | 46.9% | 31.9% |
| LMArena Coding | 1268 | 1152 |
| BigCodeBench Complete | 57.5% | 36.9% |
| SciCode | 26% | — |
| WeirdML | 14.4% | — |
| LiveBench Coding | 36.6% | — |
| HumanEval+ | — | 56.7% |
| MBPP+ | — | 54.8% |
Agentic & Tool Use Not comparable
Llama-3.3-70B-Instruct: 25.6 (#100), Llama 3-8B: —
| Benchmark | Llama-3.3-70B-Instruct | Llama 3-8B |
|---|---|---|
| Berkeley Function Calling Leaderboard | 31.9% | — |
| BALROG | 23% | — |
Reasoning Too close to call
Llama-3.3-70B-Instruct: 13.7 (#325), Llama 3-8B: 13.8 (#321)
| Benchmark | Llama-3.3-70B-Instruct | Llama 3-8B |
|---|---|---|
| LMArena Hard Prompts | 1257 | 1133 |
| DTBench | 59.5% | 43.9% |
| Epoch Capabilities Index | 127.32 | 116.33 |
| ForecastBench | 58.6 | 58.6 |
| SimpleBench | 19.9% | — |
| CritPt | 0% | — |
| Chess Puzzles | — | 0% |
| LiveBench Reasoning | 50.8% | — |
| LiveBench Data Analysis | 49.5% | — |
| LMCA | 17.5% | — |
| Adversarial NLI | — | 57.3% |
| LiveBench | 50.2% | — |
| WinoGrande | — | 75.7% |
Math Llama-3.3-70B-Instruct leads
Llama-3.3-70B-Instruct: 14.3 (#292), Llama 3-8B: 8.3 (#317)
| Benchmark | Llama-3.3-70B-Instruct | Llama 3-8B |
|---|---|---|
| OTIS Mock AIME 2024-2025 | 5.1% | 1.9% |
| LMArena Math | 1268 | 1151 |
| MATH Level 5 | 41.6% | 6.1% |
| LiveBench Math | 42.2% | — |
Knowledge Llama-3.3-70B-Instruct leads
Llama-3.3-70B-Instruct: 30.5 (#220), Llama 3-8B: 7.8 (#303)
| Benchmark | Llama-3.3-70B-Instruct | Llama 3-8B |
|---|---|---|
| GPQA Diamond | 47.4% | 26.1% |
| LMArena Expert | 1225 | 1113 |
| MMLU | 86.3% | 68.8% |
| Confabulations | 22.8% | — |
| Vectara Hallucination Rate | 4.1% | — |
| ARC (AI2) Challenge | — | 82.8% |
| OpenBookQA | — | 82.6% |
| TriviaQA | — | 67.7% |
Multilingual Llama-3.3-70B-Instruct leads
Llama-3.3-70B-Instruct: 40.1 (#215), Llama 3-8B: 31.0 (#256)
| Benchmark | Llama-3.3-70B-Instruct | Llama 3-8B |
|---|---|---|
| LMArena Non-English | 1236 | 1098 |
| LMArena Chinese | 1217 | 1076 |
| LMArena French | 1280 | 1158 |
| LMArena German | 1251 | 1104 |
| LMArena Japanese | 1149 | 967 |
| LMArena Korean | 1142 | 1003 |
| LMArena Russian | 1252 | 1109 |
| LMArena Spanish | 1269 | 1172 |
Instruction Following Llama-3.3-70B-Instruct leads
Llama-3.3-70B-Instruct: 71.3 (#152), Llama 3-8B: 58.7 (#255)
| Benchmark | Llama-3.3-70B-Instruct | Llama 3-8B |
|---|---|---|
| LMArena Instruction Following | 1242 | 1127 |
| LiveBench Instruction Following | 82.7% | — |
Long Context Llama 3-8B leads
Llama-3.3-70B-Instruct: 26.4 (#290), Llama 3-8B: 34.2 (#246)
| Benchmark | Llama-3.3-70B-Instruct | Llama 3-8B |
|---|---|---|
| LMArena Longer Query | 1256 | 1128 |
| Fiction.LiveBench | 33.3% | — |
Writing & Preference Llama-3.3-70B-Instruct leads
Llama-3.3-70B-Instruct: 47.8 (#203), Llama 3-8B: 37.7 (#251)
| Benchmark | Llama-3.3-70B-Instruct | Llama 3-8B |
|---|---|---|
| LMArena Text | 1274 | 1166 |
| LMArena Creative Writing | 1250 | 1150 |
| LMArena Multi-Turn | 1279 | 1152 |
| LiveBench Language | 39.2% | — |
Frequently asked questions
Is Llama-3.3-70B-Instruct better than Llama 3-8B?
Llama-3.3-70B-Instruct is the stronger model overall, scoring 30.3 to 25.3 on the Noometry Index.
Which is cheaper, Llama-3.3-70B-Instruct 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-3.3-70B-Instruct lists at $0.10 and $0.32.
Is Llama-3.3-70B-Instruct or Llama 3-8B better for coding?
They score almost the same on coding (30.6 vs 30.5); test both on your own repository before choosing.
Which has the bigger context window?
Llama-3.3-70B-Instruct does, with 128K tokens against 8K.
How many benchmarks do Llama-3.3-70B-Instruct and Llama 3-8B share?
26 benchmarks have published results for both models. Llama-3.3-70B-Instruct has 43 scored results on Noometry and Llama 3-8B has 34.