GPT-4 vs Llama 3.1 Instruct 405B
Wins 0 of 4 areas
—
Wins 3 of 4 areas
Reasoning · Facts · Following instructions
Llama 3.1 Instruct 405B is the stronger all-rounder.
Scores updated · 20 tests both models report · How we compare
Where each one wins
Tests won in each of the four areas where both have results. Each piece is one test, so a longer bar means more evidence; grey means the two scored within a point of each other.
- FactsGetting facts right instead of making them up02Llama 3.1 Instruct 405B2 of 2 tests
- ReasoningHard problems that need careful thinking01Llama 3.1 Instruct 405B1 of 1 test
- Following instructionsDoing exactly what it is asked01Llama 3.1 Instruct 405B1 of 1 test
- AgentsCarrying out multi-step tasks on its own00Even0 each · 1 tie
Agents, reasoning and following instructions rest on a single test each.
What it costs
Prices per million tokens, roughly 750,000 words. The bars show the cost of a million tokens read plus a million written.
The biggest differences
The tests each model wins by the widest margin, up to three each. Scores are out of 100.
Where Llama 3.1 Instruct 405B pulls ahead
- Avoids making up answers it doesn't knowAA-Omniscience · Non-hallucination+17points ahead
- Graduate-level biology, physics and chemistry questionsGPQA Diamond+16.6points ahead
- Follows unfamiliar, precisely checkable instructionsIFBench+5.8points ahead
Where GPT-4 pulls ahead
No clear win on a test scored out of 100.
Every test, side by side
All 20 tests both models report. The winning score is in its model's colour; marks a score checked independently.
ReasoningLlama 3.1 Instruct 405B
- GPQA DiamondLlama 3.1 Instruct 405B by 16.634.951.5+16.6
FactsLlama 3.1 Instruct 405B
- AA-Omniscience · Non-hallucinationLlama 3.1 Instruct 405B by 1730.647.6+17
- AA-Omniscience · AccuracyLlama 3.1 Instruct 405B by 2.22123.2+2.2
Following instructionsLlama 3.1 Instruct 405B
- IFBenchLlama 3.1 Instruct 405B by 5.833.239+5.8
Other results15 tests, not counted
Tests outside the eight areas. They are not counted above: several are summary scores built from other tests, or the same test under another name.
- MATHLlama 3.1 Instruct 405B by 31.342.573.8+31.3
- HumanEvalLlama 3.1 Instruct 405B by 226789+22
- MGSMLlama 3.1 Instruct 405B by 17.174.591.6+17.1
- AA-OmniscienceLlama 3.1 Instruct 405B by 16.7-33.8-17.1+16.7
- MMLULlama 3.1 Instruct 405B by 15.173.588.6+15.1
- HellaSwagGPT-4 by 6.195.389.2+6.1
- AIR-Bench 2024GPT-4 by 5.664.258.6+5.6
- DROPLlama 3.1 Instruct 405B by 3.980.984.8+3.9
- GSM8KLlama 3.1 Instruct 405B by 3.693.296.8+3.6
- WinoGrandeGPT-4 by 2.387.585.2+2.3
- OpenBookQAGPT-4 by 29694+2
- NarrativeQAGPT-4 by 1.976.874.9+1.9
- Artificial Analysis Coding IndexLlama 3.1 Instruct 405B by 1.413.114.5+1.4
- AA Intelligencetie6.77.3tie
- naturalquestions_closedbooktie45.745.6tie
Questions people ask
Which is better, GPT-4 or Llama 3.1 Instruct 405B?
Llama 3.1 Instruct 405B wins three of the four areas where both have results: reasoning, facts and following instructions. GPT-4 wins none. They are level on agents.
How do you compare the two?
We use the 20 benchmark tests both models have published scores on. The verdict counts the 5 tests in the eight capability areas, and a gap under one point (ten on rating-style scales) counts as a tie. The other 15 are listed but not counted, because several are summary scores or repeat a test. Each score is the one shown on the model's own page.