Claude 3 Haiku vs Llama 3.1 8B Instruct
Wins 3 of 6 areas
Coding · Long documents · Following instructions
Wins 0 of 6 areas
—
Claude 3 Haiku wins more areas, narrowly.Llama 3.1 8B Instruct is cheaper.
Scores updated · 28 tests both models report · How we compare
Where each one wins
Tests won in each of the six 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.
- CodingWriting and fixing software10Claude 3 Haiku1 of 2 tests · 1 tie
- Long documentsFinding answers in very long texts10Claude 3 Haiku1 of 1 test
- Following instructionsDoing exactly what it is asked10Claude 3 Haiku1 of 1 test
- AgentsCarrying out multi-step tasks on its own00Even0 each · 1 tie
- ReasoningHard problems that need careful thinking11Even1 each · 1 tie
- FactsGetting facts right instead of making them up11Even1 each
Agents, long documents 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 8B Instruct costs 95% 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 Claude 3 Haiku pulls ahead
- Graduate-level biology, physics and chemistry questionsGPQA Diamond+11.5points ahead
- Reasons across sets of long documentsAA-LCR+9.7points ahead
- Answers hard knowledge questions correctlyAA-Omniscience · Accuracy+9.1points ahead
Where Llama 3.1 8B Instruct pulls ahead
- Avoids making up answers it doesn't knowAA-Omniscience · Non-hallucination+37.5points ahead
- Very hard expert questions across many subjectsHumanity's Last Exam+1.2points ahead
Every test, side by side
All 28 tests both models report. The winning score is in its model's colour; marks a score checked independently.
CodingClaude 3 Haiku
- SciCodeClaude 3 Haiku by 5.418.613.2+5.4
- Terminal-Bench Hardtie0.80.8tie
ReasoningEven
- GPQA DiamondClaude 3 Haiku by 11.537.425.9+11.5
- Humanity's Last ExamLlama 3.1 8B Instruct by 1.24.15.3+1.2
- CritPttie00tie
FactsEven
- AA-Omniscience · Non-hallucinationLlama 3.1 8B Instruct by 37.519.557+37.5
- AA-Omniscience · AccuracyClaude 3 Haiku by 9.117.68.5+9.1
Following instructionsClaude 3 Haiku
- IFBenchClaude 3 Haiku by 7.536.128.6+7.5
Other results18 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.
- NarrativeQALlama 3.1 8B Instruct by 51.224.475.6+51.2
- HarmBenchClaude 3 Haiku by 29.791.361.6+29.7
- AIR-Bench 2024Claude 3 Haiku by 20.482.762.3+20.4
- DROPClaude 3 Haiku by 18.978.459.5+18.9
- AA-OmniscienceLlama 3.1 8B Instruct by 17.7-48.6-30.9+17.7
- XSTestLlama 3.1 8B Instruct by 1085.395.3+10
- OpenBookQAClaude 3 Haiku by 9.883.874+9.8
- GSM8KClaude 3 Haiku by 9.188.979.8+9.1
- MMLULlama 3.1 8B Instruct by 6.866.273+6.8
- naturalquestions_closedbookLlama 3.1 8B Instruct by 6.514.420.9+6.5
- MGSMClaude 3 Haiku by 6.275.168.9+6.2
- ARC-ChallengeClaude 3 Haiku by 5.889.283.4+5.8
- τ²-Bench Telecom (AA run)Claude 3 Haiku by 4.721.116.4+4.7
- HumanEvalClaude 3 Haiku by 3.375.972.6+3.3
- anthropic_red_teamClaude 3 Haiku by 3.210096.8+3.2
- AA IntelligenceLlama 3.1 8B Instruct by 1.35.66.9+1.3
- Artificial Analysis Coding IndexClaude 3 Haiku by 1.36.75.4+1.3
- simple_safety_testsClaude 3 Haiku by 1.210098.8+1.2
Questions people ask
Which is better, Claude 3 Haiku or Llama 3.1 8B Instruct?
Claude 3 Haiku wins three of the six areas where both have results: coding, long documents and following instructions. Llama 3.1 8B Instruct wins none, but costs 95% less. They are level on agents, reasoning and facts.
Which is better for coding?
Claude 3 Haiku. It wins 1 of the 2 coding tests both models report; Llama 3.1 8B Instruct wins none, and 1 is a tie.
Which is cheaper?
Claude 3 Haiku costs $0.25 per million input tokens and $1.25 per million output tokens; Llama 3.1 8B Instruct costs $0.02 and $0.05. That makes Llama 3.1 8B Instruct about 95% cheaper for the same work.
How do you compare the two?
We use the 28 benchmark tests both models have published scores on. The verdict counts the 10 tests in the eight capability areas, and a gap under one point (ten on rating-style scales) counts as a tie. The other 18 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.