Claude 3 Haiku vs Llama 3.2 Instruct 11B (Vision)
Wins 5 of 6 areas
Coding · Facts · Images and charts · Long documents · Following instructions
Wins 0 of 6 areas
—
Claude 3 Haiku is the stronger all-rounder.Llama 3.2 Instruct 11B (Vision) is cheaper.
Scores updated · 20 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.
- FactsGetting facts right instead of making them up20Claude 3 Haiku2 of 2 tests
- CodingWriting and fixing software10Claude 3 Haiku1 of 2 tests · 1 tie
- Images and chartsUnderstanding pictures, charts and video10Claude 3 Haiku1 of 1 test
- 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
Images and charts, 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 3.2 Instruct 11B (Vision) costs 54% 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+15.3points ahead
- Reasons across sets of long documentsAA-LCR+15points ahead
- Code for real scientific research problemsSciCode+7.4points ahead
Where Llama 3.2 Instruct 11B (Vision) pulls ahead
- Very hard expert questions across many subjectsHumanity's Last Exam+1.4points ahead
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.
CodingClaude 3 Haiku
- SciCodeClaude 3 Haiku by 7.418.611.2+7.4
- Terminal-Bench Hardtie0.80.8tie
ReasoningEven
- GPQA DiamondClaude 3 Haiku by 15.337.422.1+15.3
- Humanity's Last ExamLlama 3.2 Instruct 11B (Vision) by 1.44.15.5+1.4
- CritPttie00tie
FactsClaude 3 Haiku
- AA-Omniscience · AccuracyClaude 3 Haiku by 717.610.6+7
- AA-Omniscience · Non-hallucinationClaude 3 Haiku by 1.419.518.1+1.4
Images and chartsClaude 3 Haiku
- MMMU-ProClaude 3 Haiku by 1.530.829.3+1.5
Following instructionsClaude 3 Haiku
- IFBenchClaude 3 Haiku by 5.736.130.4+5.7
Other results10 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 3.2 Instruct 11B (Vision) by 51.224.475.6+51.2
- MMLUClaude 3 Haiku by 15.166.251.1+15.1
- AA-OmniscienceClaude 3 Haiku by 14-48.6-62.6+14
- OpenBookQAClaude 3 Haiku by 11.483.872.4+11.4
- naturalquestions_closedbookLlama 3.2 Instruct 11B (Vision) by 914.423.4+9
- GSM8KClaude 3 Haiku by 6.688.982.3+6.6
- τ²-Bench Telecom (AA run)Claude 3 Haiku by 6.521.114.6+6.5
- Artificial Analysis Coding IndexClaude 3 Haiku by 2.56.74.2+2.5
- AA Agentic IndexClaude 3 Haiku by 2.174.9+2.1
- AA Intelligencetie5.65.4tie
Questions people ask
Which is better, Claude 3 Haiku or Llama 3.2 Instruct 11B (Vision)?
Claude 3 Haiku wins five of the six areas where both have results: coding, facts, images and charts, long documents and following instructions. Llama 3.2 Instruct 11B (Vision) wins none, but costs 54% 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 3.2 Instruct 11B (Vision) 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 3.2 Instruct 11B (Vision) costs $0.34 and $0.34. That makes Llama 3.2 Instruct 11B (Vision) about 54% cheaper for the same work.
How do you compare the two?
We use the 20 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 10 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.