Llama 3.3 70B
Llama 3.3 70B places in the top three on PHYSICS.
44
Benchmarks with a score
0
Ranked first, out of five or more models
1
In the top three
Meta Dec 2024 Llama 3.3 70B
Open weights
126.1
#170 of 256
Epoch Capabilities Index
- GPQA Diamond 47.4% #167 /229
- KoBALT 32% #4 /12
- MolecularIQ 8.97% #8 /17
Where it ranks
Its best score on each benchmark, and where that ranks among every model with a score there. We rank each model once, at its best setting.
Knowledge and language
| Benchmark | Score | Rank | Source | |
|---|---|---|---|---|
| MMLU | 86.3% | #4 of 85 | Claim in Phi-4 Technical Report, via Epoch AI | |
| KoBALT | 32% | #4 of 12 | Official leaderboard | |
| SnorkelSpatial | 14.6% | #18 of 27 | Official leaderboard | |
| LongBench v2 | 36.2% | #26 of 33 | Official leaderboard | |
| LLM Confabulation Leaderboard | 22.8% | #46 of 58 | Official leaderboard | |
| Fiction.LiveBench | 33.3% | #54 of 57 | Fiction.live leaderboard, via Epoch AI |
Math and science
| Benchmark | Score | Rank | Source | |
|---|---|---|---|---|
| PHYSICS | 31.5% | #3 of 20 | Official leaderboard | |
| MATH Level 5 | 41.6% | #60 of 94 | Epoch AI run | |
| SciPredict | 18.2% | #8 of 12 | Official leaderboard | |
| GPQA Diamond | 47.4% | #167 of 229 | Epoch AI run | |
| DTBench | 59.5% | #147 of 180 | Official leaderboard | |
| SnorkelGraph | 23.5% | #20 of 24 | Official leaderboard | |
| SnorkelSequences | 38.8% | #21 of 25 | Official leaderboard | |
| OTIS Mock AIME 2024-2025 | 5.14% | #181 of 212 | Epoch AI run |
Coding
| Benchmark | Score | Rank | Source | |
|---|---|---|---|---|
| BigO(Bench) | 33.7% | #12 of 18 | Official leaderboard | |
| MMTU | 45.4% | #6 of 9 | Official leaderboard | |
| SWE-PolyBench | 6.02% | #4 of 6 | Official leaderboard | |
| GitTaskBench | 20.4% | #8 of 10 | Official leaderboard |
Agents and tools
| Benchmark | Score | Rank | Source | |
|---|---|---|---|---|
| BALROG | 23% | #19 of 33 | Official leaderboard | |
| Elimination Game | 4 | #36 of 56 | Official leaderboard | |
| Factorio Learning Environment | 54998.0 | #4 of 6 | Official leaderboard | |
| Step Race | 1 | #42 of 59 | Official leaderboard | |
| LLM Chess | -608 | #113 of 123 | Official leaderboard | |
| SnorkelWordle | 10.2% | #19 of 19 | Official leaderboard |
Professional work
| Benchmark | Score | Rank | Source | |
|---|---|---|---|---|
| TheAgentCompany | 6.90% | #4 of 7 | Official leaderboard | |
| SnorkelUnderwrite | 46.3% | #18 of 27 | Official leaderboard | |
| LegalBench | 77.2% | #101 of 138 | Official leaderboard | |
| FinanceReasoning | 80.7% | #5 of 6 | Official leaderboard | |
| TaxEval v2 | 59.4% | #117 of 129 | Official leaderboard | |
| Vals Finance Agent v1 | 3.36% | #45 of 47 | Official leaderboard | |
| StrataBench Land v1 Extract | 85.5% | #5 of 5 | Official leaderboard | |
| FinEval 6.0 | 77.3% | #4 of 4 | Official leaderboard |
Safety and security
| Benchmark | Score | Rank | Source | |
|---|---|---|---|---|
| MolecularIQ | 8.97% | #8 of 17 | Official leaderboard | |
| BioTIER Permit | 99.4% | #33 of 69 | Official leaderboard | |
| FORTRESS | 44.8% | #54 of 73 | Official leaderboard | |
| BaxBench | 10.7% | #26 of 34 | Official leaderboard | |
| BioTIER Refuse | 21.2% | #56 of 69 | Official leaderboard | |
| AutoBaxBench | 5.00% | #10 of 11 | Official leaderboard | |
| Human Pathogen Capabilities Test v2 (HPCT v2) | 24.8% | #68 of 71 | Official leaderboard | |
| AIRTBench | 0% | #9 of 9 | Official leaderboard |
Other
| Benchmark | Score | Rank | Source | |
|---|---|---|---|---|
| Political Science LLM Benchmark | 66.9% | #17 of 42 | Official leaderboard | |
| ACCoRD | 45.5 | #120 of 152 | Official leaderboard | |
| LMCA | 17.5% | #118 of 147 | Official leaderboard | |
| Conceptual Reasoning Index | 26.1 | #124 of 151 | Official leaderboard |
Meta's models over time
Meta's models in color, against every other model on the Epoch Capabilities Index, by release date.
Source: Epoch Capabilities Index, Epoch AI (CC BY 4.0)