The Open-Source AI Stack
RSS
All models

Models · Compare

Qwen 2.5 72B Instruct vs Claude 3.5 Haiku

Rows highlighted in warm gray are where the models differ. Numbers carry their as-of date and primary source.

Specs

Field A: Qwen 2.5 72B Instruct B: Claude 3.5 Haiku
Released 2024-09-192024-11-04
Developer AlibabaAnthropic
Openness OpenProprietary
License Qwen License (Apache-2.0 for 72B variant requires agreement)Proprietary
OSI-approved nono
Data released nono
Training code nono
Architecture denseunknown
Total params 72.7B
Active params
Experts
Context window 131K200K
Attention gqaunknown
Position enc. ropeunknown
Pretraining tokens 18.0T
Post-training sft, dporlhf, constitutional
Training hardware
$/M input $0.36$0.80
$/M output $0.40$4.00
Output tok/sec 54.10

Benchmarks

Missing scores render as not reported; never inferred. Bold highlights the leader per benchmark.

General reasoning

MMLU-Pro 72.0 2026-05-21 63.4 2026-05-21
GPQA-Diamond 49.0 2024-09-19 40.8 2026-05-21

Code

HumanEval 86.6 2024-09-19
SWE-Bench Verified 40.6 2024-10-22
LiveCodeBench 27.6 2026-05-21 31.4 2026-05-21

Math

MATH 85.8 2026-05-21 72.1 2026-05-21
AIME 2024 16.0 2026-05-21 3.3 2026-05-21
AIME 2025 14.0 2026-05-21

Context · A

The 72B class set new open-weights leadership across most reasoning benchmarks at release, with an 18T-token pretrain. Released alongside coder, math, and 0.5B/1.5B/3B/7B/14B/32B sibling sizes the same day.

Context · B

Announced October 22 2024 alongside the upgraded Claude 3.5 Sonnet and the computer-use beta; generally available in early November. Anthropic positioned it as matching Claude 3 Opus on many intelligence benchmarks at Haiku-tier speeds, with a SWE-Bench Verified score of 40.6% making it briefly the best small-tier coding model from a major lab.

Qwen 2.5 72B Instruct detail → · Claude 3.5 Haiku detail →