Skip to content

Alibaba · Chat

Qwen2.5 72B Instruct

Alibaba 🇨🇳ChatQwen familyOne of its sources says the weights aren't publicly available.The maker lists this model as deprecated (29 Jul 2026).
From NanoGPT's listing:
Great multilingual support, strong at mathematics and coding, supports roleplay and chatbots.
llminfo IQ
108.9
Market median price
$0.37 blended / 1M
Max context
131K
Listings
26

Evidence and sources

International storefront (alibabacloud.com) is Singapore-incorporated; mainland operating entity 杭州阿里云智能科技有限公司 (Hangzhou).

🤗 Qwen/Qwen2.5-72B-Instructllminfo IQ 108.9, ranked among chat models. From 216 benchmarks. Plenty of data behind it.How llminfo IQ is computedCoding IQ 79.9, 95% interval 58–102. From 9 benchmark families across 6 sources. A fair amount of data behind it. Uses the model's best published results. It has its own scale, separate from llminfo IQ.How Coding IQ is computedPoor value (35.0). Costs more per task than its benchmark results predict. From 7 benchmarks with measured task costs. A fair amount of data behind it.How Value for money worksReleased 19 Sep 2024 Training-data cutoff, as its sources state it Maker HQ: CN (source)
All rankings: #299 on OpenRouter today #302 on OpenRouter this week #318 on OpenRouter this month #584 on NanoGPT recentlySources: LMArena data, DeepInfra data, LiteLLM data, Models.dev data, OpenRouter data docs, NanoGPT, llminfo researchPrimary data from LMArena from DeepInfra: name from LiteLLM: deprecation date from Models.dev: weights availability from OpenRouter: knowledge cutoff from NanoGPT: description from llminfo research: release date

Advantages and trade-offs

Compared with 10 chat models, picked mainly by popularity and IQ

Trade-offs

Deprecated by its maker

The maker has already deprecated this model (29 Jul 2026).

No open weights

The weights aren't public, so you can't run it yourself.

We only make a claim when there's enough data on models of the same type. Updated 22 Nov, 06:07 UTC.Prices come from the model's maker where they sell it, otherwise their official partner, otherwise the market median. This comparison uses 7 maker's prices; 3 market median prices.

Input & output

as stated by the model's providers
Text
Text

Capabilities

Supported operations and provider-stated features

Function callingStructured output

Listings

 OpenRouterqwen/qwen-2.5-72b-instruct:free · freeSource: Helicone Costs dataNot published by the provider$0.00$0.00
 AbacusQwen/Qwen2.5-72B-InstructSource: Models.dev data128K$0.11$0.38
 Hyperbolichyperbolic/Qwen/Qwen2.5-72B-InstructSource: LiteLLM data131K$0.12$0.30
 Nebius AI Studionebius/Qwen/Qwen2.5-72B-InstructSource: LiteLLM data128K$0.13$0.40
 RequestyInference by DeepInfradeepinfra/Qwen/Qwen2.5-72B-InstructSource: Requesty data131K$0.23$0.40
 FastrouterQwen/Qwen2.5-72B-InstructSource: FastRouter data33K$0.36$0.40
 FlowdotQwen/Qwen2.5-72B-InstructSource: FlowDot data33K$0.36$0.40
 Hugging FaceInference by DeepInfraQwen/Qwen2.5-72B-InstructSource: Hugging Face Inference Providers data33K$0.36$0.40
 Kiloqwen/qwen-2.5-72b-instructSources: Kilo AI Gateway data, Models.dev data33K$0.36$0.40
 NanoGPTqwen/qwen-2.5-72b-instructSources: Models.dev data, NanoGPT131K$0.36cache read $0.18$0.40
Show all 25 listings

Showing 10 of 25: the cheapest per 1M input tokens from each provider.

Also sold by 302ai, Aihubmix, DeepInfra, Fireworks AI, Nous Portal, Novita AI, Openroute Cn, SiliconFlow, Siliconflow Cn, Together AI.

1 listing without a price is hidden.

Benchmarks

Named effort levels run from lowest to highest. Select a result to see its source details.

Coding

CodeElo28 related benchmarks · 28 results
HumanEvalSource: LLM Stats data · Reported by the model's developer. The source relaying it did not verify it. claim
MBPPSource: LLM Stats data · Reported by the model's developer. The source relaying it did not verify it. claim
Multi-SWE-bench9 related benchmarks · 27 results
C Verified
Source: Multi-SWE-bench Experiments data docs · Mar 2025
MagentLess, 20250329_MagentLess_Qwen2.5-72B-Instruct
MopenHands, 20250329_MopenHands_Qwen2.5-72B-Instruct
MSWE-agent, 20250329_MSWE-agent_Qwen2.5-72B-Instruct
C++ Verified
Source: Multi-SWE-bench Experiments data docs · Mar 2025
MagentLess, 20250329_MagentLess_Qwen2.5-72B-Instruct
MopenHands, 20250329_MopenHands_Qwen2.5-72B-Instruct
MSWE-agent, 20250329_MSWE-agent_Qwen2.5-72B-Instruct
Go Verified
Source: Multi-SWE-bench Experiments data docs · Mar 2025
MagentLess, 20250329_MagentLess_Qwen2.5-72B-Instruct
MopenHands, 20250329_MopenHands_Qwen2.5-72B-Instruct
MSWE-agent, 20250329_MSWE-agent_Qwen2.5-72B-Instruct
Java Verified
Source: Multi-SWE-bench Experiments data docs · Mar 2025
MagentLess, 20250329_MagentLess_Qwen2.5-72B-Instruct
MopenHands, 20250329_MopenHands_Qwen2.5-72B-Instruct
MSWE-agent, 20250329_MSWE-agent_Qwen2.5-72B-Instruct
JavaScript Verified
Source: Multi-SWE-bench Experiments data docs · Mar 2025
MagentLess, 20250329_MagentLess_Qwen2.5-72B-Instruct
MopenHands, 20250329_MopenHands_Qwen2.5-72B-Instruct
MSWE-agent, 20250329_MSWE-agent_Qwen2.5-72B-Instruct
Lite All Verified
Source: Multi-SWE-bench Experiments data docs · Mar 2025
MagentLess, 20250329_MagentLess_Qwen2.5-72B-Instruct
MopenHands, 20250329_MopenHands_Qwen2.5-72B-Instruct
MSWE-agent, 20250329_MSWE-agent_Qwen2.5-72B-Instruct
Show 9 more resultsHide 9 results
Python Verified
Source: Multi-SWE-bench Experiments data docs · Mar 2025
MagentLess, 20250329_Agentless_Qwen2.5-72B-Instruct
MopenHands, 20250329_OpenHands_Qwen2.5-72B-Instruct
MSWE-agent, 20250329_SWE-agent_Qwen2.5-72B-Instruct
Rust Verified
Source: Multi-SWE-bench Experiments data docs · Mar 2025
MagentLess, 20250329_MagentLess_Qwen2.5-72B-Instruct
MopenHands, 20250329_MopenHands_Qwen2.5-72B-Instruct
MSWE-agent, 20250329_MSWE-agent_Qwen2.5-72B-Instruct
TypeScript Verified
Source: Multi-SWE-bench Experiments data docs · Mar 2025
MagentLess, 20250329_MagentLess_Qwen2.5-72B-Instruct
MopenHands, 20250329_MopenHands_Qwen2.5-72B-Instruct
MSWE-agent, 20250329_MSWE-agent_Qwen2.5-72B-Instruct
SWE-rebenchSource: SWE-rebench

Math

GSM8KSource: LLM Stats data · Reported by the model's developer. The source relaying it did not verify it. claim
LiveBenchSource: LLM Stats data · Reported by the model's developer. The source relaying it did not verify it. claim
MATHSource: LLM Stats data · Reported by the model's developer. The source relaying it did not verify it. claim
mu-MATH7 related benchmarks · 35 results
F1
Source: U-MATH and mu-MATH
F1 · Qwen/Qwen2.5-72B-Instruct
accuracy
Source: U-MATH and mu-MATH
accuracy · Qwen/Qwen2.5-72B-Instruct
MCC
Source: U-MATH and mu-MATH
MCC · Qwen/Qwen2.5-72B-Instruct
NPV
Source: U-MATH and mu-MATH
NPV · Qwen/Qwen2.5-72B-Instruct
PPV
Source: U-MATH and mu-MATH
PPV · Qwen/Qwen2.5-72B-Instruct
TNR
Source: U-MATH and mu-MATH
TNR · Qwen/Qwen2.5-72B-Instruct
Show 5 more resultsHide 5 results
TPR
Source: U-MATH and mu-MATH
TPR · Qwen/Qwen2.5-72B-Instruct
U-MATH7 related benchmarks · 21 results
algebra accuracy
Source: U-MATH and mu-MATH
Task set · algebra · gpt-4o-2024-08-06
differential_calc accuracy
Source: U-MATH and mu-MATH
Task set · differential_calc · gpt-4o-2024-08-06
integral_calc accuracy
Source: U-MATH and mu-MATH
Task set · integral_calc · gpt-4o-2024-08-06
multivariable_calculus accuracy
Source: U-MATH and mu-MATH
Task set · multivariable_calculus · gpt-4o-2024-08-06
overall accuracy
Source: U-MATH and mu-MATH
Task set · u_math · gpt-4o-2024-08-06
precalculus_review accuracy
Source: U-MATH and mu-MATH
Task set · precalculus_review · gpt-4o-2024-08-06
Show 3 more resultsHide 3 results
sequences_series accuracy
Source: U-MATH and mu-MATH
Task set · sequences_series · gpt-4o-2024-08-06

Human preference

Arena ELO28 related benchmarks · 30 results
Arena ELO
Source: LMArena data · Nov 2026
Arena Hard
Source: LLM Stats data · Reported by the model's developer. The source relaying it did not verify it. claim
Arena-Hard-v2.0-Preview
Source: Arena-Hard-Auto data
Creative Writing, Ensemble GPT-4.1 and Gemini 2.5 as Judges (Best Configuration for Creative Writing)
Hard Prompt, Style Control, and Gemini-2.5 as Judge (Official Configuration)
Hard Prompt, Style Control, and GPT-4.1 as Judge (If prefer OpenAI API)
Chinese
Source: LMArena data · Nov 2026
Creative Writing
Source: LMArena data · Nov 2026
English
Source: LMArena data · Nov 2026
Show 22 more resultsHide 22 results
Exclude Ties
Source: LMArena data · Nov 2026
Expert
Source: LMArena data · Nov 2026
French
Source: LMArena data · Nov 2026
German
Source: LMArena data · Nov 2026
Hard Prompts
Source: LMArena data · Nov 2026
Hard Prompts English
Source: LMArena data · Nov 2026
Industry Mathematical
Source: LMArena data · Nov 2026
Japanese
Source: LMArena data · Nov 2026
Korean
Source: LMArena data · Nov 2026
Longer Query
Source: LMArena data · Nov 2026
Multi Turn
Source: LMArena data · Nov 2026
Non English
Source: LMArena data · Nov 2026
Russian
Source: LMArena data · Nov 2026
Spanish
Source: LMArena data · Nov 2026
Style-Controlled
Source: LMArena data · Nov 2026
MT-BenchSource: LLM Stats data · Reported by the model's developer. The source relaying it did not verify it. claim

Multilingual

Agentic

Source: OpenRouter benchmark exports
qwen-2.5-72b-instruct · 90-day results
Across 2 providers lowest 25% median 26.2% highest 27.3%OpenRouter's automatic routing 24%
Show results from 2 providers and OpenRouter's automatic routingHide results from 2 providers and OpenRouter's automatic routing
  • Novita
  • DeepInfra
  • OpenRouter's automatic routing

Instruction following

Factuality

LLM-AggreFact12 related benchmarks · 12 results
CNN
Source: LLM-AggreFact data
Qwen2.5-72B-Instruct, grounded claim checker, CNN, renderer-v1
all-11 mean
Source: LLM-AggreFact data
Qwen2.5-72B-Instruct, grounded claim checker, all-11 mean, renderer-v1
ClaimVerify
Source: LLM-AggreFact data
Qwen2.5-72B-Instruct, grounded claim checker, ClaimVerify, renderer-v1
ExpertQA
Source: LLM-AggreFact data
Qwen2.5-72B-Instruct, grounded claim checker, ExpertQA, renderer-v1
FactCheck
Source: LLM-AggreFact data
Qwen2.5-72B-Instruct, grounded claim checker, FactCheck, renderer-v1
LFQA
Source: LLM-AggreFact data
Qwen2.5-72B-Instruct, grounded claim checker, LFQA, renderer-v1
Show 6 more resultsHide 6 results
MediaS
Source: LLM-AggreFact data
Qwen2.5-72B-Instruct, grounded claim checker, MediaS, renderer-v1
MeetB
Source: LLM-AggreFact data
Qwen2.5-72B-Instruct, grounded claim checker, MeetB, renderer-v1
RAGTruth
Source: LLM-AggreFact data
Qwen2.5-72B-Instruct, grounded claim checker, RAGTruth, renderer-v1
REVEAL
Source: LLM-AggreFact data
Qwen2.5-72B-Instruct, grounded claim checker, REVEAL, renderer-v1
WiCE
Source: LLM-AggreFact data
Qwen2.5-72B-Instruct, grounded claim checker, WiCE, renderer-v1
XSum
Source: LLM-AggreFact data
Qwen2.5-72B-Instruct, grounded claim checker, XSum, renderer-v1

General

MultiPL-ESource: LLM Stats data · Reported by the model's developer. The source relaying it did not verify it. claim

Reasoning

Safety

Science

GPQASource: LLM Stats data · Reported by the model's developer. The source relaying it did not verify it. claim
GPQA Diamond4 results
Source: Epoch AI Benchmarking Hub data
Setting not stated
Source: OpenRouter benchmark exports · Oct 2026
qwen-2.5-72b-instruct · 90-day results
Across 2 providers lowest 46.1% median 46.9% highest 47.7%OpenRouter's automatic routing 46%
Show results from 2 providers and OpenRouter's automatic routingHide results from 2 providers and OpenRouter's automatic routing
  • DeepInfra
  • Novita
  • OpenRouter's automatic routing

Knowledge

Vision

Other

Open Greek FinLLM11 related benchmarks · 11 results
QA raw score
Source: TheFinAI public numeric results
greek-contents, QA, QA Raw
FinNum raw score
Source: TheFinAI public numeric results
greek-contents, FinNum, FinNum Raw
FinText raw score
Source: TheFinAI public numeric results
greek-contents, FinText, FinText Raw
Show 5 more resultsHide 5 results
FNS raw score
Source: TheFinAI public numeric results
greek-contents, FNS, FNS Raw
MultiFin raw score
Source: TheFinAI public numeric results
greek-contents, MultiFin, MultiFin Raw
overall normalized index
Source: TheFinAI public numeric results
greek-contents, published-average

Data credits: ACEBench, MIT (licence) · Data by LLM Stats · LMArena, CC BY 4.0 (licence) · Arena-Hard-Auto, LMArena, Apache-2.0 (licence) · ArtifactsBench (Tencent Hunyuan), CC BY 4.0 (licence) · Epoch AI Benchmarking Hub, CC BY 4.0 (licence) · EQ-Bench (Samuel J. Paech) — cite paech2023eqbench · CodeElo (Qwen). Aggregate scores from the official results file; no results licence asserted. The evaluation scaffold agreement is not presented as a licence for these scores · EvalPlus Leaderboard and EvalPerf (evalplus.github.io), Apache-2.0; EvalPerf win-rate fractions converted to percentages (licence) · FaithJudge, Vectara; Tamber et al., EMNLP 2025 Industry. Public numeric README leaderboard; results licence not separately established. · Galileo Agent Leaderboard, Apache-2.0 (licence) · OpenRouter public benchmark leaderboard and search-lane exports, CC BY 4.0. Fractional source scores converted to percentages; immutable dated snapshots. (licence) · KazBench by Yersultan Akhmer. Provisional DEV numeric results; code MIT and task-data CC BY 4.0 do not establish a separate score licence. · LiveMCPBench leaderboard, ICIP-CAS (Mo et al., 2025). Sheet linked from the Apache-2.0 project README; sheet licence not separately stated. · LLM Chess by Maxim Saplin, Apache-2.0 (Copyright [2024] [Maxim Saplin]) (licence) · LLM-AggreFact, Tang et al. Public numeric grounded checker balanced accuracy; results licence not separately established. · Toloka, U-MATH and mu-MATH public leaderboard. MIT Space; generated solution outputs retain underlying model terms and are not ingested. (licence) · Multi-SWE-bench experiments (multi-swe-bench/experiments), Apache-2.0; resolved-rate fractions converted to percentages (licence) · Open Arabic LLM Leaderboard (OALL), Apache-2.0 (licence) · The Fin AI. Published FinBen result summaries and Open Greek Financial LLM Leaderboard aggregates. Results licence not separately stated; fractional FinBen success metrics converted to percentages. · PashtoBench v1.0, Jawad Noori. Public results/scores numbers; code MIT and data/ CC BY 4.0 grants do not name results/, so no results licence is asserted. · pfgen-bench, Preferred Networks. Apache-2.0 except LLM-generated outputs; aggregate scores only. (licence) · SeaExam and SeaBench leaderboard, SeaLLMs (DAMO-NLP-SG). Published leaderboard scores; results licence not stated. · SnakeBench (Greg Kamradt / ARC Prize Foundation), MIT (licence) · WebWalkerQA (Alibaba NLP), Apache-2.0 (licence)

Also known as: deepinfra/qwen/qwen2.5-72b-instruct, fireworks_ai/accounts/fireworks/models/qwen2p5-72b-instruct, hyperbolic/qwen/qwen2.5-72b-instruct, nebius/qwen/qwen2.5-72b-instruct, novita/qwen/qwen-2.5-72b-instruct, qwen/qwen-2.5-72b-instruct, qwen/qwen-2.5-72b-instruct:free, qwen/qwen2.5-72b-instruct, sophnet/qwen2.5-72b-instruct, together_ai/qwen/qwen2.5-72b-instruct

Like this one. Tap or hover any dotted value, label or icon to see where it comes from and how it is measured. open details when tapped or hovered. How this data works