Enter opens the highlighted result. Esc closes.

MiniMax M3 vs Qwen3.8 27B

Qwen3.8 27B leads MiniMax M3 on coding, 74.2 to 39.2 on the Vibe Coding Index, and MiniMax M3 costs 2.1x less per output token ($1.20 vs $2.55 per million). Updated October 22, 2026; 80 benchmarks have a result for both.

Replace MiniMax M3 Replace Qwen3.8 27B
MiniMax M3MiniMax
Coding
39.2
Overall
51.7
Output $/M
$1.20
21
Qwen3.8 27BQwen
Coding
74.2
Overall
59.6
Output $/M
$2.55
59
vs boards won of 80 shared

MiniMax M3 or Qwen3.8 27B: which to pick

Pick Qwen3.8 27B for coding: it scores 74.2 on the Vibe Coding Index against 39.2 for MiniMax M3, though on 4 benchmarks against 6.

Pick MiniMax M3 for output tokens 2.1x cheaper ($1.20 vs $2.55 per million) and a lower cost per solved task ($0.56 against $1.26).

Model AMiniMax M3MiniMax39.2/1006 independent benchmarks$0.56 per solved taskModel BQwen3.8 27BQwen74.2/1004 independent benchmarks$1.26 per solved task
Shared results
80
MiniMax M3 only
37
Qwen3.8 27B only
5
Like-for-like categories
8 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
MiniMax M3 39.2 from 6 benchmarks, Qwen3.8 27B 74.2 from 4 benchmarks.
Widest category
Agentic work, 11.4 points apart on average over 9 shared benchmarks.
Widest single gap
LiveBench browser: MiniMax M3 8.0% against Qwen3.8 27B 32.0%, 24.0 points apart on a 0 to 100 scale.
Output price
MiniMax M3 $1.20 against Qwen3.8 27B $2.55 per million output tokens, 2.1x apart.
Context window
MiniMax M3 takes 1.05M tokens, Qwen3.8 27B 1M.

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
CategoryMiniMax M3BarsQwen3.8 27BTies
Coding313
Intelligence02
Agentic work18
Reasoning37
Math12
Knowledge54
Web design01
General822
All shared boards2159
CodingLike-for-likeSix of these boards feed the Vibe Coding Index313
Coding benchmarks for MiniMax M3 and Qwen3.8 27B
BenchmarkMiniMax M3Qwen3.8 27BLeader
LiveBench coding70.9%86.8%Qwen3.8 27B
LMArena WebDev14821593Qwen3.8 27B
AA Coding Index58.668.1Qwen3.8 27B
LiveCodeBench (Vals AI)82.2%84.0%Qwen3.8 27B
SWE-bench Verified (Vals AI)75.0%86.0%Qwen3.8 27B
Terminal-Bench 4 (Vals AI)1.0%Not measuredOnly MiniMax M3 measured
Vibe Code Bench (Vals AI)47.6%64.8%Qwen3.8 27B
Code migration (Vals AI)19.9%14.2%MiniMax M3
ProgramBench (Vals AI)Not measured0.0%Only Qwen3.8 27B measured
IOI (Vals AI)Not measured39.1%Only Qwen3.8 27B measured
Terminal-Bench 2.1 (Vals AI)53.6%58.4%Qwen3.8 27B
SWE-bench Multimodal (Vals AI)33.3%Not measuredOnly MiniMax M3 measured
Vibe Code Bench 1-100 (Vals AI)9.2%Not measuredOnly MiniMax M3 measured
SkillsBench (Vals AI)51.5%38.1%MiniMax M3
CyberBench (Vals AI)65.0%67.8%Qwen3.8 27B
SciCode47.1%46.6%MiniMax M3
Terminal-Bench Hard (Artificial Analysis run)42.4%Not measuredOnly MiniMax M3 measured
FrontierCode14.7%Not measuredOnly MiniMax M3 measured
ALE-Bench640.0Not measuredOnly MiniMax M3 measured
LMArena Text: coding14701483Qwen3.8 27B
LMArena Text: software and IT14661478Qwen3.8 27B
LMArena WebDev: React14781602Qwen3.8 27B
LMArena WebDev: HTML14551570Qwen3.8 27B
LMArena WebDev: from an image14701577Qwen3.8 27B
Neither model has a result on the other 19 boards here
IntelligenceLike-for-likePublic boards, shown as published, kept out of the index02
Intelligence benchmarks for MiniMax M3 and Qwen3.8 27B
BenchmarkMiniMax M3Qwen3.8 27BLeader
AA Intelligence Index29.233.7Qwen3.8 27B
Epoch Capabilities Index146.9151.7Qwen3.8 27B
Agentic workLike-for-likePublic boards, shown as published, kept out of the index18
Agentic work benchmarks for MiniMax M3 and Qwen3.8 27B
BenchmarkMiniMax M3Qwen3.8 27BLeader
AA Agentic Index29.545.8Qwen3.8 27B
Design Arena: agents building web apps1200Not measuredOnly MiniMax M3 measured
Design Arena: agents building full-stack apps1197Not measuredOnly MiniMax M3 measured
Design Arena: agents building mobile apps1169Not measuredOnly MiniMax M3 measured
Design Arena: agents building Android apps1158Not measuredOnly MiniMax M3 measured
Design Arena: agents building games1159Not measuredOnly MiniMax M3 measured
Design Arena: HTML slides1120Not measuredOnly MiniMax M3 measured
Design Arena: PowerPoint slides1178Not measuredOnly MiniMax M3 measured
Design Arena: spreadsheets1042Not measuredOnly MiniMax M3 measured
LiveBench research68.0%59.4%MiniMax M3
LiveBench browser8.0%32.0%Qwen3.8 27B
LiveBench long agentic12.4%27.9%Qwen3.8 27B
Finance Agent (Vals AI)48.3%48.5%Qwen3.8 27B
Tax Agent Bench (Vals AI)20.6%30.9%Qwen3.8 27B
Legal Research Bench (Vals AI)29.8%36.1%Qwen3.8 27B
Excel Modeling Benchmark (Vals AI)47.8%59.7%Qwen3.8 27B
Terminal-Bench-Science (Vals AI)Not measured0.0%Only Qwen3.8 27B measured
τ²-Bench Telecom88.9%Not measuredOnly MiniMax M3 measured
APEX-Agents37.7%47.5%Qwen3.8 27B
OSWorld 2.04.6%Not measuredOnly MiniMax M3 measured
Vending-Bench 2$2,158Not measuredOnly MiniMax M3 measured
Neither model has a result on the other 21 boards here
ReasoningLike-for-likePublic boards, shown as published, kept out of the index37
Reasoning benchmarks for MiniMax M3 and Qwen3.8 27B
BenchmarkMiniMax M3Qwen3.8 27BLeader
GPQA Diamond92.7%88.9%MiniMax M3
LiveBench reasoning74.5%80.0%Qwen3.8 27B
MMMU Pro (Vals AI)81.2%83.9%Qwen3.8 27B
GPQA Diamond (Artificial Analysis run)92.9%90.5%MiniMax M3
Humanity's Last Exam (Artificial Analysis run)39.0%33.9%MiniMax M3
GPQA Diamond (Epoch AI run)90.9%Not measuredOnly MiniMax M3 measured
ARC-AGINot measured87.5%Only Qwen3.8 27B measured
ARC-AGI-2Not measured42.4%Only Qwen3.8 27B measured
SimpleBench45.8%Not measuredOnly MiniMax M3 measured
Chess puzzles13.0%Not measuredOnly MiniMax M3 measured
Mystery game puzzles8.0%Not measuredOnly MiniMax M3 measured
LMArena Text: hard prompts14471459Qwen3.8 27B
LMArena Text: expert14611479Qwen3.8 27B
LMArena Text: hard prompts in English14551461Qwen3.8 27B
LMArena Vision: diagrams12661285Qwen3.8 27B
LMArena Vision: homework12681295Qwen3.8 27B
Neither model has a result on the other 7 boards here
MathLike-for-likePublic boards, shown as published, kept out of the index12
Math benchmarks for MiniMax M3 and Qwen3.8 27B
BenchmarkMiniMax M3Qwen3.8 27BLeader
OTIS Mock AIME71.1%Not measuredOnly MiniMax M3 measured
ProofBench18.0%16.0%MiniMax M3
LMArena Text: math14291455Qwen3.8 27B
LMArena Text: mathematical work14481458Qwen3.8 27B
Neither model has a result on the other 10 boards here
KnowledgeLike-for-likePublic boards, shown as published, kept out of the index54
Knowledge benchmarks for MiniMax M3 and Qwen3.8 27B
BenchmarkMiniMax M3Qwen3.8 27BLeader
MMLU Pro84.2%84.3%Qwen3.8 27B
LegalBench (Vals AI)85.4%82.4%MiniMax M3
CorpFin (Vals AI)68.1%Not measuredOnly MiniMax M3 measured
TaxEval (Vals AI)72.7%70.8%MiniMax M3
MortgageTax (Vals AI)68.4%64.9%MiniMax M3
MedCode (Vals AI)46.3%28.7%MiniMax M3
LMArena Text: science14521467Qwen3.8 27B
LMArena Text: medicine and healthcare14411437MiniMax M3
LMArena Text: legal and government14371447Qwen3.8 27B
LMArena Text: business and finance14341438Qwen3.8 27B
Neither model has a result on the other 10 boards here
Web designLike-for-likePublic boards, shown as published, kept out of the index01
Web design benchmarks for MiniMax M3 and Qwen3.8 27B
BenchmarkMiniMax M3Qwen3.8 27BLeader
Design Arena: websites1257Not measuredOnly MiniMax M3 measured
Design Arena: UI components1243Not measuredOnly MiniMax M3 measured
Design Arena: game dev1210Not measuredOnly MiniMax M3 measured
Design Arena: data viz1236Not measuredOnly MiniMax M3 measured
Design Arena: 3D1208Not measuredOnly MiniMax M3 measured
Design Arena: all code categories1246Not measuredOnly MiniMax M3 measured
Design Arena: SVG1160Not measuredOnly MiniMax M3 measured
Design Arena: ASCII art1161Not measuredOnly MiniMax M3 measured
LiveBench design66.2%75.9%Qwen3.8 27B
GeneralLike-for-likePublic boards, shown as published, kept out of the index822
General benchmarks for MiniMax M3 and Qwen3.8 27B
BenchmarkMiniMax M3Qwen3.8 27BLeader
LiveBench data analysis76.2%76.6%Qwen3.8 27B
Public Benefits Bench (Vals AI)64.1%Not measuredOnly MiniMax M3 measured
SAGE (Vals AI)50.6%52.4%Qwen3.8 27B
MedScribe (Vals AI)87.3%83.8%MiniMax M3
Vals Multimodal Index60.0%Not measuredOnly MiniMax M3 measured
IFBench82.9%Not measuredOnly MiniMax M3 measured
Long context reasoning (AA-LCR)83.0%82.0%MiniMax M3
DTBench78.9%88.0%Qwen3.8 27B
LMCA33.7%41.4%Qwen3.8 27B
Surface Evolver Bench55.0%45.0%MiniMax M3
LMArena Text14331440Qwen3.8 27B
LMArena Text: instruction following14331436Qwen3.8 27B
LMArena Text: creative writing14041383MiniMax M3
LMArena Vision12511269Qwen3.8 27B
LMArena Text: English14431448Qwen3.8 27B
LMArena Text: non-English14201428Qwen3.8 27B
LMArena Text: Chinese14661501Qwen3.8 27B
LMArena Text: German14281443Qwen3.8 27B
LMArena Text: Japanese13801387Qwen3.8 27B
LMArena Text: Korean13731384Qwen3.8 27B
LMArena Text: French14441465Qwen3.8 27B
LMArena Text: Spanish14331449Qwen3.8 27B
LMArena Text: Russian14261415MiniMax M3
LMArena Text: Polish14231439Qwen3.8 27B
LMArena Text: multi-turn14431440MiniMax M3
LMArena Text: longer queries14461448Qwen3.8 27B
LMArena Text: writing and language14191405MiniMax M3
LMArena Text: entertainment and media14031393MiniMax M3
LMArena Vision: English12631267Qwen3.8 27B
LMArena Vision: Chinese13111326Qwen3.8 27B
LMArena Vision: reading text12591281Qwen3.8 27B
LMArena Vision: humor12351235Qwen3.8 27B
LMArena Vision: recognizing things1197Not measuredOnly MiniMax M3 measured
LMArena Vision: creative writing12341237Qwen3.8 27B
LMArena Document1435Not measuredOnly MiniMax M3 measured
EQ-Bench 41150Not measuredOnly MiniMax M3 measured
Neither model has a result on the other 12 boards here
Image generationNot comparablePublic boards, shown as published, kept out of the index0 boards, none shared
Image generation benchmarks for MiniMax M3 and Qwen3.8 27B
BenchmarkMiniMax M3Qwen3.8 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

MiniMax M3$0.0009Fits in one request
Qwen3.8 27B$0.0017Fits in one request

MiniMax M3 costs less

Repository review

50K input + 3K output tokens

MiniMax M3$0.019Fits in one request
Qwen3.8 27B$0.029Fits in one request

MiniMax M3 costs less

Agent loop

200K input + 10K output tokens

MiniMax M3$0.072Fits in one request
Qwen3.8 27B$0.111Fits in one request

MiniMax M3 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 MiniMax M3 and Qwen3.8 27B
SpecificationMiniMax M3Qwen3.8 27B
MakerMiniMaxQwen
WeightsOpen weightsOpen weights
ReleasedMay 31, 2026August 14, 2026
Context window1.05M1M
Input $/Mtok$0.30$0.425
Output $/Mtok$1.20$2.55
Coding index39.274.2
Independent measurements behind it64
Cost per solved task$0.56$1.26
SourceOpenRouter listingOpenRouter listing

Questions

Is MiniMax M3 better than Qwen3.8 27B for coding?

Pick Qwen3.8 27B for coding: it scores 74.2 on the Vibe Coding Index against 39.2 for MiniMax M3, though on 4 benchmarks against 6. Pick MiniMax M3 for output tokens 2.1x cheaper ($1.20 vs $2.55 per million) and a lower cost per solved task ($0.56 against $1.26).

Which is cheaper, MiniMax M3 or Qwen3.8 27B?

MiniMax M3 costs $0.30 per million input tokens and $1.20 per million output tokens; Qwen3.8 27B costs $0.425 and $2.55. Per solved SWE-bench Verified task, MiniMax M3 costs $0.56 and Qwen3.8 27B $1.26.

Which has the bigger context window, MiniMax M3 or Qwen3.8 27B?

MiniMax M3, with 1.05M tokens against 1M.

How many benchmarks is the MiniMax M3 vs Qwen3.8 27B comparison based on?

80 benchmarks have a result for both models as of October 22, 2026. MiniMax M3's coding index rests on 6 independent benchmarks and Qwen3.8 27B's on 4. Where only one model has a result, the page says so instead of naming a winner.

Compare next

Embed this comparison on your site

Free under CC BY 4.0, while the source link stays visible. Preview the frame · Terms and other embeds

<iframe src="https://llmbenchmark.ai/embed/compare/minimax-minimax-m3-vs-qwen3-8-27b" title="MiniMax M3 vs Qwen3.8 27B, LLMBenchmark.ai" width="100%" height="360" style="border:0;max-width:100%" loading="lazy"></iframe>
<p style="font:13px/1.4 sans-serif;margin:6px 0 0">Source: <a href="https://llmbenchmark.ai/compare/minimax-minimax-m3-vs-qwen3-8-27b">LLMBenchmark.ai, MiniMax M3 vs Qwen3.8 27B</a></p>