Llama 3
Llama 3 is an open-weights AI model from Meta with an index score of 101 (GPT-4 = 100), rank 194 of 217.
10184–118Index score#194 of 217 models
–Price per million tokensno public API price found
–Context windowtokens it reads at once
18 Apr 2024ReleasedReplaced by Llama 3.1-70B
Where it sitson the frontier
GPT-4 (Jun 2023) reached this score 10 months earlier.
Test resultsby area · tick = best by any model
Reasoning & knowledgeevidence
Mathsevidence
Codingno evidence yet
Agentsevidence
Professional workno evidence yet
Novel problemsno evidence yet
Long tasksno evidence yet
Third-party results only, at the model’s best standard setting. Hover a bar for who ran it.
What others sayrank on 8 leaderboards
The Model Indexthird-party tests, our method#194of 217 · 101
LMArena · Creative writingcreative writing#243of 360 · 1254Epoch Capabilities Indexoverall capability from many benchmarks#160of 231 · 121.3LMArena · Mathsmaths questions in chat#240of 346 · 1259LMArena · Textoverall chat quality, judged by people#255of 362 · 1277LMArena · Instruction followingfollowing detailed instructions#255of 362 · 1259LMArena · Codingcoding questions in chat#255of 357 · 1306LMArena · Long promptslong, complex requests#271of 340 · 1254Best attop three among current models
Replaced by Llama 3.1-70B; it no longer leads any category among current models.
Facts and sources
Released18 Apr 2024
Announcementai.meta.com ↗
Release entryLlama 3
Scored asLlama 3-70B
Output speed–
Hugging Facenot linked