Claude 3 Haiku vs Llama 3 Instruct 8B
Wins 4 of 5 areas
Coding · Facts · Long documents · Following instructions
Wins 0 of 5 areas
—
Claude 3 Haiku is the stronger all-rounder.Llama 3 Instruct 8B is cheaper.
Scores updated · 28 tests both models report · How we compare
Where each one wins
Tests won in each of the five 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 up20Claude 3 Haiku2 of 2 tests
- 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
- ReasoningHard problems that need careful thinking11Even1 each · 1 tie
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 Instruct 8B costs 87% 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
- Follows unfamiliar, precisely checkable instructionsIFBench+11.5points ahead
- Avoids making up answers it doesn't knowAA-Omniscience · Non-hallucination+9.3points ahead
- Graduate-level biology, physics and chemistry questionsGPQA Diamond+7.8points ahead
Where Llama 3 Instruct 8B pulls ahead
- Very hard expert questions across many subjectsHumanity's Last Exam+1points 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 6.718.611.9+6.7
- Terminal-Bench Hardtie0.80tie
ReasoningEven
- GPQA DiamondClaude 3 Haiku by 7.837.429.6+7.8
- Humanity's Last ExamLlama 3 Instruct 8B by 14.15.1+1
- CritPttie00tie
FactsClaude 3 Haiku
- AA-Omniscience · Non-hallucinationClaude 3 Haiku by 9.319.510.2+9.3
- AA-Omniscience · AccuracyClaude 3 Haiku by 7.217.610.4+7.2
Following instructionsClaude 3 Haiku
- IFBenchClaude 3 Haiku by 11.536.124.6+11.5
Other results19 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 Instruct 8B by 5124.475.4+51
- GSM8KClaude 3 Haiku by 3988.949.9+39
- naturalquestions_closedbookLlama 3 Instruct 8B by 23.414.437.8+23.4
- AA-OmniscienceClaude 3 Haiku by 21.5-48.6-70.1+21.5
- τ²-Bench Telecom (AA run)Claude 3 Haiku by 21.121.10+21.1
- HarmBenchClaude 3 Haiku by 18.691.372.7+18.6
- BBHClaude 3 Haiku by 1673.757.7+16
- HumanEvalClaude 3 Haiku by 15.575.960.4+15.5
- HellaSwagClaude 3 Haiku by 14.885.971.1+14.8
- AIR-Bench 2024Claude 3 Haiku by 11.882.770.9+11.8
- XSTestLlama 3 Instruct 8B by 10.385.395.6+10.3
- OpenBookQAClaude 3 Haiku by 7.283.876.6+7.2
- AA Agentic IndexClaude 3 Haiku by 770+7
- ARC-ChallengeClaude 3 Haiku by 6.489.282.8+6.4
- MMLUClaude 3 Haiku by 666.260.1+6
- Artificial Analysis Coding IndexClaude 3 Haiku by 2.76.74+2.7
- anthropic_red_teamClaude 3 Haiku by 1.210098.8+1.2
- AA Intelligencetie5.64.8tie
- simple_safety_teststie10099.3tie
Questions people ask
Which is better, Claude 3 Haiku or Llama 3 Instruct 8B?
Claude 3 Haiku wins four of the five areas where both have results: coding, facts, long documents and following instructions. Llama 3 Instruct 8B wins none, but costs 87% less. They are level on reasoning.
Which is better for coding?
Claude 3 Haiku. It wins 1 of the 2 coding tests both models report; Llama 3 Instruct 8B 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 Instruct 8B costs $0.04 and $0.14. That makes Llama 3 Instruct 8B about 87% 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 9 tests in the eight capability areas, and a gap under one point (ten on rating-style scales) counts as a tie. The other 19 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.