Enter opens the highlighted result. Esc closes.

Qwen2.5 72B Instruct vs Qwen3.6 27B

Qwen3.6 27B leads Qwen2.5 72B Instruct on the Artificial Analysis Intelligence Index, 21.4 to 7.7, and Qwen2.5 72B Instruct costs 6.8x less per output token ($0.40 vs $2.70 per million). Updated October 22, 2026; 11 benchmarks have a result for both.

Replace Qwen2.5 72B Instruct Replace Qwen3.6 27B
Qwen2.5 72B InstructQwen
Coding
No data
Overall
26.5
Output $/M
$0.40
0
Qwen3.6 27BQwen
Coding
36.8
Overall
45.3
Output $/M
$2.70
11
vs boards won of 11 shared

Qwen2.5 72B Instruct or Qwen3.6 27B: which to pick

Only Qwen3.6 27B has a coding score, 36.8 on the Vibe Coding Index; Qwen2.5 72B Instruct has none, so the two cannot be ranked on coding.

Pick Qwen2.5 72B Instruct for output tokens 6.8x cheaper ($0.40 vs $2.70 per million).

Model AQwen2.5 72B InstructQwenNot benchmarkedNo coding benchmarkNo cost per solved taskModel BQwen3.6 27BQwen36.8/1002 independent benchmarks$1.30 per solved task
Shared results
11
Qwen2.5 72B Instruct only
42
Qwen3.6 27B only
13
Like-for-like categories
6 of 9

Only benchmarks that tested both models count toward a pick; a result only one model has is shown but not compared.

Which one for your work

Each card is decided by one number both models carry. Where only one has it, the card says so and makes no call.

At a glance

Coding index
Qwen2.5 72B Instruct not benchmarked, Qwen3.6 27B 36.8 from 2 benchmarks.
Widest category
Math, 83.1 points apart on average over 1 shared benchmark.
Widest single gap
OTIS Mock AIME: Qwen2.5 72B Instruct 8.1% against Qwen3.6 27B 91.1%, 83.1 points apart on a 0 to 100 scale.
Output price
Qwen2.5 72B Instruct $0.40 against Qwen3.6 27B $2.70 per million output tokens, 6.8x apart.
Context window
Qwen2.5 72B Instruct takes 33K tokens, Qwen3.6 27B 262K.

Benchmark by benchmark

First the score: boards each model leads, per category, where both have a result. Then every board, a category to a fold; bars are each result on a 0 to 100 scale, so compare the two in a row, not one row with another.

Boards won, where both have a result
CategoryQwen2.5 72B InstructBarsQwen3.6 27BTies
Coding01
Intelligence02
Agentic work01
Reasoning03
Math01
General03
All shared boards011
CodingLike-for-likeSix of these boards feed the Vibe Coding Index01
Coding benchmarks for Qwen2.5 72B Instruct and Qwen3.6 27B
BenchmarkQwen2.5 72B InstructQwen3.6 27BLeader
AA Coding IndexNot measured53.7Only Qwen3.6 27B measured
SWE-bench Verified (Vals AI)Not measured70.0%Only Qwen3.6 27B measured
Vibe Code Bench (Vals AI)Not measured11.9%Only Qwen3.6 27B measured
LiveCodeBench (Artificial Analysis run)27.6%Not measuredOnly Qwen2.5 72B Instruct measured
SciCodeNot measured42.8%Only Qwen3.6 27B measured
Terminal-Bench Hard (Artificial Analysis run)4.5%34.8%Qwen3.6 27B
WeirdML16.0%Not measuredOnly Qwen2.5 72B Instruct measured
LMArena Text: coding1292Not measuredOnly Qwen2.5 72B Instruct measured
LMArena Text: software and IT1292Not measuredOnly Qwen2.5 72B Instruct measured
Neither model has a result on the other 30 boards here
IntelligenceLike-for-likePublic boards, shown as published, kept out of the index02
Intelligence benchmarks for Qwen2.5 72B Instruct and Qwen3.6 27B
BenchmarkQwen2.5 72B InstructQwen3.6 27BLeader
AA Intelligence Index7.721.4Qwen3.6 27B
Epoch Capabilities Index129.0146.2Qwen3.6 27B
Agentic workLike-for-likePublic boards, shown as published, kept out of the index01
Agentic work benchmarks for Qwen2.5 72B Instruct and Qwen3.6 27B
BenchmarkQwen2.5 72B InstructQwen3.6 27BLeader
AA Agentic IndexNot measured18.5Only Qwen3.6 27B measured
τ²-Bench Telecom34.5%94.2%Qwen3.6 27B
METR time horizons35.8%Not measuredOnly Qwen2.5 72B Instruct measured
BALROG16.2%Not measuredOnly Qwen2.5 72B Instruct measured
TheAgentCompany5.7%Not measuredOnly Qwen2.5 72B Instruct measured
Neither model has a result on the other 36 boards here
ReasoningLike-for-likePublic boards, shown as published, kept out of the index03
Reasoning benchmarks for Qwen2.5 72B Instruct and Qwen3.6 27B
BenchmarkQwen2.5 72B InstructQwen3.6 27BLeader
GPQA Diamond (Artificial Analysis run)49.1%84.2%Qwen3.6 27B
Humanity's Last Exam (Artificial Analysis run)3.6%23.1%Qwen3.6 27B
GPQA Diamond (Epoch AI run)49.1%85.9%Qwen3.6 27B
Mystery game puzzlesNot measured7.0%Only Qwen3.6 27B measured
BIG-Bench Hard79.8%Not measuredOnly Qwen2.5 72B Instruct measured
LMArena Text: hard prompts1271Not measuredOnly Qwen2.5 72B Instruct measured
LMArena Text: expert1245Not measuredOnly Qwen2.5 72B Instruct measured
LMArena Text: hard prompts in English1279Not measuredOnly Qwen2.5 72B Instruct measured
Neither model has a result on the other 14 boards here
MathLike-for-likePublic boards, shown as published, kept out of the index01
Math benchmarks for Qwen2.5 72B Instruct and Qwen3.6 27B
BenchmarkQwen2.5 72B InstructQwen3.6 27BLeader
AA Math Index14.0Not measuredOnly Qwen2.5 72B Instruct measured
AIME 2025 (Artificial Analysis run)14.0%Not measuredOnly Qwen2.5 72B Instruct measured
OTIS Mock AIME8.1%91.1%Qwen3.6 27B
FrontierMath, tiers 1 to 3Not measured35.1%Only Qwen3.6 27B measured
MATH level 563.2%Not measuredOnly Qwen2.5 72B Instruct measured
LMArena Text: math1283Not measuredOnly Qwen2.5 72B Instruct measured
LMArena Text: mathematical work1290Not measuredOnly Qwen2.5 72B Instruct measured
Neither model has a result on the other 6 boards here
KnowledgeDirectional onlyPublic boards, shown as published, kept out of the index11 boards, none shared
Knowledge benchmarks for Qwen2.5 72B Instruct and Qwen3.6 27B
BenchmarkQwen2.5 72B InstructQwen3.6 27BLeader
CaseLaw (Vals AI)Not measured53.2%Only Qwen3.6 27B measured
CorpFin (Vals AI)Not measured62.3%Only Qwen3.6 27B measured
TaxEval (Vals AI)Not measured71.3%Only Qwen3.6 27B measured
MortgageTax (Vals AI)Not measured68.3%Only Qwen3.6 27B measured
MMLU Pro (Artificial Analysis run)72.0%Not measuredOnly Qwen2.5 72B Instruct measured
MMLU85.3%Not measuredOnly Qwen2.5 72B Instruct measured
AI2 Reasoning Challenge94.5%Not measuredOnly Qwen2.5 72B Instruct measured
LMArena Text: science1275Not measuredOnly Qwen2.5 72B Instruct measured
LMArena Text: medicine and healthcare1258Not measuredOnly Qwen2.5 72B Instruct measured
LMArena Text: legal and government1288Not measuredOnly Qwen2.5 72B Instruct measured
LMArena Text: business and finance1253Not measuredOnly Qwen2.5 72B Instruct measured
Neither model has a result on the other 7 boards here
Web designNot comparablePublic boards, shown as published, kept out of the index0 boards, none shared
Web design benchmarks for Qwen2.5 72B Instruct and Qwen3.6 27B
BenchmarkQwen2.5 72B InstructQwen3.6 27BLeader
Neither model has a result on 9 boards here
GeneralLike-for-likePublic boards, shown as published, kept out of the index03
General benchmarks for Qwen2.5 72B Instruct and Qwen3.6 27B
BenchmarkQwen2.5 72B InstructQwen3.6 27BLeader
SAGE (Vals AI)Not measured45.6%Only Qwen3.6 27B measured
IFBench36.9%67.6%Qwen3.6 27B
Long context reasoning (AA-LCR)Not measured77.3%Only Qwen3.6 27B measured
DTBench62.9%78.1%Qwen3.6 27B
LMCA13.4%34.5%Qwen3.6 27B
HellaSwag84.8%Not measuredOnly Qwen2.5 72B Instruct measured
WinoGrande82.3%Not measuredOnly Qwen2.5 72B Instruct measured
PIQA82.6%Not measuredOnly Qwen2.5 72B Instruct measured
LMArena Text1269Not measuredOnly Qwen2.5 72B Instruct measured
LMArena Text: instruction following1254Not measuredOnly Qwen2.5 72B Instruct measured
LMArena Text: creative writing1221Not measuredOnly Qwen2.5 72B Instruct measured
LMArena Text: English1283Not measuredOnly Qwen2.5 72B Instruct measured
LMArena Text: non-English1252Not measuredOnly Qwen2.5 72B Instruct measured
LMArena Text: Chinese1273Not measuredOnly Qwen2.5 72B Instruct measured
LMArena Text: German1234Not measuredOnly Qwen2.5 72B Instruct measured
LMArena Text: Japanese1180Not measuredOnly Qwen2.5 72B Instruct measured
LMArena Text: Korean1188Not measuredOnly Qwen2.5 72B Instruct measured
LMArena Text: French1280Not measuredOnly Qwen2.5 72B Instruct measured
LMArena Text: Spanish1255Not measuredOnly Qwen2.5 72B Instruct measured
LMArena Text: Russian1264Not measuredOnly Qwen2.5 72B Instruct measured
LMArena Text: multi-turn1272Not measuredOnly Qwen2.5 72B Instruct measured
LMArena Text: longer queries1281Not measuredOnly Qwen2.5 72B Instruct measured
LMArena Text: writing and language1247Not measuredOnly Qwen2.5 72B Instruct measured
LMArena Text: entertainment and media1211Not measuredOnly Qwen2.5 72B Instruct measured
Neither model has a result on the other 22 boards here
Image generationNot comparablePublic boards, shown as published, kept out of the index0 boards, none shared
Image generation benchmarks for Qwen2.5 72B Instruct and Qwen3.6 27B
BenchmarkQwen2.5 72B InstructQwen3.6 27BLeader
Neither model has a result on 5 boards here

What each workload costs

Three fixed token mixes at each model's listed price, so per-token rates become a bill.

Chat turn

1K input + 500 output tokens

Qwen2.5 72B Instruct$0.0006Fits in one request
Qwen3.6 27B$0.0018Fits in one request

Qwen2.5 72B Instruct costs less

Repository review

50K input + 3K output tokens

Qwen2.5 72B Instruct$0.019Exceeds the context window
Qwen3.6 27B$0.031Fits in one request

Qwen2.5 72B Instruct costs less

Agent loop

200K input + 10K output tokens

Qwen2.5 72B Instruct$0.076Exceeds the context window
Qwen3.6 27B$0.117Fits in one request

Qwen2.5 72B Instruct costs less

Listed OpenRouter rates, without cache or batch discounts. A workload that does not fit a model's context window would need splitting.

Specification differences

The leading value in a row is marked; a row with a missing value names no leader. Our two rows are marked in red.

Specifications of Qwen2.5 72B Instruct and Qwen3.6 27B
SpecificationQwen2.5 72B InstructQwen3.6 27B
MakerQwenQwen
WeightsOpen weightsOpen weights
ReleasedSeptember 19, 2024April 27, 2026
Context window33K262K
Input $/Mtok$0.36$0.45
Output $/Mtok$0.40$2.70
Coding indexNot benchmarked36.8
Independent measurements behind it02
Cost per solved taskNot computable$1.30
SourceOpenRouter listingOpenRouter listing

Questions

Is Qwen2.5 72B Instruct better than Qwen3.6 27B for coding?

Only Qwen3.6 27B has a coding score, 36.8 on the Vibe Coding Index; Qwen2.5 72B Instruct has none, so the two cannot be ranked on coding. Pick Qwen2.5 72B Instruct for output tokens 6.8x cheaper ($0.40 vs $2.70 per million).

Which is cheaper, Qwen2.5 72B Instruct or Qwen3.6 27B?

Qwen2.5 72B Instruct costs $0.36 per million input tokens and $0.40 per million output tokens; Qwen3.6 27B costs $0.45 and $2.70. Cost per solved task is known for Qwen3.6 27B only.

Which has the bigger context window, Qwen2.5 72B Instruct or Qwen3.6 27B?

Qwen3.6 27B, with 262K tokens against 33K.

How many benchmarks is the Qwen2.5 72B Instruct vs Qwen3.6 27B comparison based on?

11 benchmarks have a result for both models as of October 22, 2026. Qwen2.5 72B Instruct's coding index rests on 0 independent benchmarks and Qwen3.6 27B's on 2. Where only one model has a result, the page says so instead of naming a winner.

Compare next