Models/Qwen2.5 Instruct 72B

MDL

Qwen2.5 Instruct 72B

Alibaba

Verified against Artificial Analysis · Jul 13, 2026

Capability

percentile index

Price /M

$0.475 / $0.495

in / out

Context

—

Avg score

34.1

11 benchmarks

AA Intelligence

9.6

index

Weights

Open

Released

Sep 19, 2024

About

Qwen2.5 Instruct 72B is a open-weights model in the Qwen2.5 family from Alibaba. Benchmarked on 11 evals, averaging 34.1.

Benchmark scores · 11

avg 34.1 — higher is better

MATHartificial-analysis · 2026-07

85.8

MMLU-Proartificial-analysis · 2026-07

GPQA Diamondartificial-analysis · 2026-07

49.1

IFBenchartificial-analysis · 2026-07

36.9

τ²-benchartificial-analysis · 2026-07

34.5

LiveCodeBenchartificial-analysis · 2026-07

27.6

SciCodeartificial-analysis · 2026-07

26.7

AA Long-Context Reasoningartificial-analysis · 2026-07

20.3

AIMEartificial-analysis · 2026-07

Terminal-Bench Hardartificial-analysis · 2026-07

4.5

Humanity's Last Examartificial-analysis · 2026-07

4.2

The research behind this model

all papers mentioning it →

BitCal-TTS: Bit-Calibrated Test-Time Scaling for Quantized Reasoning ModelsMay 7, 2026

Mentions matched by name in title/abstract — from the arXiv + HF daily corpus.

Details

PublisherAlibaba

FamilyQwen2.5

Parameters—

License—

WeightsOpen

ReleasedSep 19, 2024

Performance & price

Price in / out$0.475 / $0.495 per 1M

Blended price$0.48 / 1M

Math index14.0

Median across API providers, via Artificial Analysis. Blended = 3:1 input:output. For reasoning models, latency includes thinking time. Methodology →

History

full ledger →

Capability over time

Append-only ledger — every observed change to this model's numbers.

API providers

Artificial Analysis$0.495/M