GPT-4 vs Llama 3.1 70B Instruct
Wins 1 of 4 areas
Facts
Wins 2 of 4 areas
Reasoning · Following instructions
Llama 3.1 70B Instruct wins more areas, narrowly.GPT-4 is better at facts.
Scores updated · 19 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.
- ReasoningHard problems that need careful thinking01Llama 3.1 70B Instruct1 of 1 test
- Following instructionsDoing exactly what it is asked01Llama 3.1 70B Instruct1 of 1 test
- FactsGetting facts right instead of making them up20GPT-42 of 2 tests
- 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.
Llama 3.1 70B Instruct costs 99% less for the same work.
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 70B Instruct pulls ahead
- Graduate-level biology, physics and chemistry questionsGPQA Diamond+6points ahead
- Follows unfamiliar, precisely checkable instructionsIFBench+1.2points ahead
Where GPT-4 pulls ahead
- Avoids making up answers it doesn't knowAA-Omniscience · Non-hallucination+8.8points ahead
- Answers hard knowledge questions correctlyAA-Omniscience · Accuracy+1.3points ahead
Every test, side by side
All 19 tests both models report. The winning score is in its model's colour; marks a score checked independently.
ReasoningLlama 3.1 70B Instruct
- GPQA DiamondLlama 3.1 70B Instruct by 634.940.9+6
FactsGPT-4
- AA-Omniscience · Non-hallucinationGPT-4 by 8.830.621.8+8.8
- AA-Omniscience · AccuracyGPT-4 by 1.32119.7+1.3
Following instructionsLlama 3.1 70B Instruct
- IFBenchLlama 3.1 70B Instruct by 1.233.234.4+1.2
Other results14 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 70B Instruct by 25.542.568+25.5
- AIR-Bench 2024GPT-4 by 21.764.242.5+21.7
- HumanEvalLlama 3.1 70B Instruct by 13.56780.5+13.5
- MMLULlama 3.1 70B Instruct by 12.573.586+12.5
- MGSMLlama 3.1 70B Instruct by 12.474.586.9+12.4
- AA-OmniscienceGPT-4 by 9.3-33.8-43.1+9.3
- Artificial Analysis Coding IndexGPT-4 by 2.213.110.9+2.2
- OpenBookQAGPT-4 by 2.29693.8+2.2
- WinoGrandeGPT-4 by 2.287.585.3+2.2
- DROPGPT-4 by 1.380.979.6+1.3
- GSM8Ktie93.293.8tie
- naturalquestions_closedbooktie45.745.2tie
- NarrativeQAtie76.877.2tie
- AA Intelligencetie6.76.6tie
Questions people ask
Which is better, GPT-4 or Llama 3.1 70B Instruct?
Llama 3.1 70B Instruct wins two of the four areas where both have results: reasoning and following instructions. GPT-4 wins facts. They are level on agents.
Which is cheaper?
GPT-4 costs $30.00 per million input tokens and $60.00 per million output tokens; Llama 3.1 70B Instruct costs $0.56 and $0.56. That makes Llama 3.1 70B Instruct about 99% cheaper for the same work.
How do you compare the two?
We use the 19 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 14 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.