Compare · ModelsLive · 2 picked · head to head

Claude 2.1 vs Claude 3.5 Haiku

Side by side · benchmarks, pricing, and signals you can act on.

Winner summary

Claude 3.5 Haiku wins 4 of 4 shared benchmarks. Leads in knowledge · math · coding.

Category leads
knowledge·Claude 3.5 Haikumath·Claude 3.5 Haikucoding·Claude 3.5 Haiku
Hype vs Reality
Claude 2.1
#213 by perf·no signal
QUIET
Claude 3.5 Haiku
#159 by perf·no signal
QUIET
Best value
Claude 2.1
no price
Claude 3.5 Haiku
15.5 pts/$
$2.40/M
Vendor risk
Anthropic logo
Anthropic
$380.0B·Tier 1
Medium risk
Anthropic logo
Anthropic
$380.0B·Tier 1
Medium risk
Head to head
Claude 2.1Claude 3.5 Haiku
GPQA diamond
Claude 3.5 Haiku leads by +6.9
Graduate-Level Google-Proof QA (Diamond set) · expert-crafted questions in physics, biology, and chemistry that are difficult even for domain PhDs.
Claude 2.1
10.6
Claude 3.5 Haiku
17.5
MMLU
Claude 3.5 Haiku leads by +1.1
Massive Multitask Language Understanding · 57 subjects spanning STEM, humanities, social sciences, and more. The standard benchmark for broad knowledge.
Claude 2.1
64.7
Claude 3.5 Haiku
65.7
OTIS Mock AIME 2024-2025
Claude 3.5 Haiku leads by +2.4
OTIS Mock AIME 2024-2025 · simulated American Invitational Mathematics Examination problems testing advanced problem-solving skills.
Claude 2.1
1.9
Claude 3.5 Haiku
4.2
WeirdML
Claude 3.5 Haiku leads by +23.7
WeirdML · tests models on unusual and adversarial machine learning tasks that require creative problem-solving beyond standard patterns.
Claude 2.1
7.1
Claude 3.5 Haiku
30.7
Full benchmark table
BenchmarkClaude 2.1Claude 3.5 Haiku
GPQA diamond
Graduate-Level Google-Proof QA (Diamond set) · expert-crafted questions in physics, biology, and chemistry that are difficult even for domain PhDs.
10.617.5
MMLU
Massive Multitask Language Understanding · 57 subjects spanning STEM, humanities, social sciences, and more. The standard benchmark for broad knowledge.
64.765.7
OTIS Mock AIME 2024-2025
OTIS Mock AIME 2024-2025 · simulated American Invitational Mathematics Examination problems testing advanced problem-solving skills.
1.94.2
WeirdML
WeirdML · tests models on unusual and adversarial machine learning tasks that require creative problem-solving beyond standard patterns.
7.130.7
Pricing · per 1M tokens · projected $/mo at 10M tokens
ModelInputOutputContextProjected $/mo
Anthropic logoClaude 2.1
Anthropic logoClaude 3.5 Haiku$0.80$4.00200K tokens (~100 books)$16.00