MiniMax M3 vs Qwen3.8 Max

Updated October 29, 2026. Rank says Qwen3.8 Max is ahead. Price, access, and your workload can each overturn that. Public scores include evidence status and uncertainty.

Decision reading

Qwen3.8 Max has the higher public score estimate, 64.86 versus 48.57, but the 90% score intervals overlap. Treat that as a lead, not a settled winner. 17 results are shared. Category rows resting on Estimated evidence or different benchmark sets are marked directional and do not name a winner.

Model A

MiniMax M3

MiniMax

48.57/100

Supported · Public rank #104

90% interval 32.6–64.5

Model B

Qwen3.8 Max

Alibaba

64.86/100

Supported · Public rank #33

90% interval 56.0–73.7

Shared results: 17
MiniMax M3 only: 11
Qwen3.8 Max only: 49
Like-for-like categories: 3 / 8

Supported: MiniMax M3 and Qwen3.8 MaxHow the comparison works

Which one for your work

Recommendations appear only when a shared evidence basis or an explicit operating constraint supports the call. Secondary and unsupported use cases stay disclosed below the initial list.

Coding work
Code generation, repair, and software-engineering tasks
Qwen3.8 Max
Qwen3.8 Max has the higher public coding point estimate, 52.7 to 37.4, with Supported evidence for both models. The conditional ranges do not establish rank confidence.
Confidence: limited
Agentic work
Tool use, computer use, and multi-step task completion
Qwen3.8 Max
Qwen3.8 Max has the higher public agentic point estimate, 66 to 39.3, with Supported evidence for both models. The conditional ranges do not establish rank confidence.
Confidence: limited

Show secondary and unsupported calls

Long documents
Prompts that approach the documented context limit
No clear pick
The documented context windows are equal.
Confidence: documented
Chat turn cost
1K fresh input + 500 output tokens
Not enough matched evidence
A complete comparable API-rate estimate is not available for both models.
Confidence: listed-rates
Cache-heavy agent loop cost
200K cached + 20K fresh input + 10K output tokens
Not enough matched evidence
A complete comparable API-rate estimate is not available for both models.
Confidence: listed-rates
Repository review cost
50K fresh input + 3K output tokens
Not enough matched evidence
A complete comparable API-rate estimate is not available for both models.
Confidence: listed-rates

Which one for a specific job

Choose a job from the LLM Selector's task catalog to see the category row it rests on, under the same basis rules as the table below. A directional row stays directional; choosing a job never creates a winner.

Pick the job you would use them for

The same task catalog as the LLM Selector. Each job names the evidence surface it rests on; nothing here adds a new score.

37.4MiniMax M352.7Qwen3.8 Max

Like-for-like · BenchAlign v5.8

Qwen3.8 Max has the higher coding point estimate. Conditional score ranges do not establish rank confidence.

Coding scores combine specific tasks and setups. Match the editor, harness, and effort to your workflow.

Open the coding leaderboard Shortlist models for this job

Same basis rules as the category table below

Shape of the matched evidence

Only shared public evidence is shown. Sparse evidence stays a ruled list rather than being closed into a radar shape.

Too few matched category axes support a radar. The ruled list below shows only shared benchmark results; positions use each benchmark’s normalized display scale when available.

OSWorld-VerifiedAgentic
Normalized gap 16.0
MiniMax M3:70.1%
Qwen3.8 Max:86.1%
MiniMax M3 source Qwen3.8 Max source
OSWorld 2.0Agentic
Normalized gap 14.8
MiniMax M3:4.6%
Qwen3.8 Max:19.4%
MiniMax M3 source Qwen3.8 Max source
SWE-bench ProCoding
Normalized gap 8.7
MiniMax M3:59%
Qwen3.8 Max:67.7%
MiniMax M3 source Qwen3.8 Max source
LiveCodeBench (Vals)Coding
Normalized gap 5.7
MiniMax M3:82.2%
Qwen3.8 Max:87.9%
MiniMax M3 source Qwen3.8 Max source
MMLU-Pro (Vals)Knowledge
Normalized gap 4.4
MiniMax M3:84.2%
Qwen3.8 Max:88.6%
MiniMax M3 source Qwen3.8 Max source

Bars run 0–100 on each benchmark’s normalized display scale

Category results, on a stated basis

Each row shows the public-lane category score for both models: the BenchAlign v5.8 lane for agentic, coding, and knowledge, and the weighted public rows elsewhere, with the same rank each model page reports. A row is like-for-like only when both scores rest on Supported evidence or the same weighted set. Directional and not-comparable rows remain visible, but they never receive a winner.

Agentic

Like-for-like

MiniMax M3: 39.3; Supported · #76/135
Qwen3.8 Max: 66.0; Supported · #16/135
Basis: BenchAlign v5.8 lane · 10 vs 18 public rows
Reading: Qwen3.8 Max leads

Coding

Like-for-like

MiniMax M3: 37.4; Supported · #75/156
Qwen3.8 Max: 52.7; Supported · #38/156
Basis: BenchAlign v5.8 lane · 10 vs 13 public rows
Reading: Qwen3.8 Max leads · intervals overlap

Knowledge

Like-for-like

MiniMax M3: 54.3; Supported · #62/179
Qwen3.8 Max: 65.5; Supported · #28/179
Basis: BenchAlign v5.8 lane · 2 vs 6 public rows
Reading: Qwen3.8 Max leads · intervals overlap

Multimodal

Directional only

MiniMax M3: 59.0; #48/58
Qwen3.8 Max: 91.9; #2/58
Basis: Provisional lane · 2 vs 2 weighted rows
Reading: Directional only

Instruction following

Directional only

MiniMax M3: 93.8; #6/157
Qwen3.8 Max: 94.7; #2/157
Basis: Provisional lane · 0 vs 1 weighted rows
Reading: Directional only

Reasoning

Not comparable

MiniMax M3: 82.0; Unranked · 2 rankable rows
Qwen3.8 Max: 87.0; Unranked · 2 rankable rows
Basis: Provisional lane · 0 vs 2 weighted rows
Reading: Not comparable

Multilingual

Not comparable

MiniMax M3: Not ranked
Qwen3.8 Max: Not ranked
Basis: Provisional lane · 0 vs 0 weighted rows
Reading: Not comparable

Math

Not comparable

MiniMax M3: Not ranked
Qwen3.8 Max: 78.9; Unranked · 2 rankable rows
Basis: Provisional lane · 1 vs 2 weighted rows
Reading: Not comparable

Category scores with the server-provided evidence basis for MiniMax M3 and Qwen3.8 Max
Category	MiniMax M3	Qwen3.8 Max	Basis	Reading
Agentic	39.3Supported · #76/135	66.0Supported · #16/135	Like-for-likeBenchAlign v5.8 lane · 10 vs 18 public rows	Qwen3.8 Max leads
Coding	37.4Supported · #75/156	52.7Supported · #38/156	Like-for-likeBenchAlign v5.8 lane · 10 vs 13 public rows	Qwen3.8 Max leads · intervals overlap
Knowledge	54.3Supported · #62/179	65.5Supported · #28/179	Like-for-likeBenchAlign v5.8 lane · 2 vs 6 public rows	Qwen3.8 Max leads · intervals overlap
Multimodal	59.0#48/58	91.9#2/58	Directional onlyProvisional lane · 2 vs 2 weighted rows	Directional only
Instruction following	93.8#6/157	94.7#2/157	Directional onlyProvisional lane · 0 vs 1 weighted rows	Directional only
Reasoning	82.0Unranked · 2 rankable rows	87.0Unranked · 2 rankable rows	Not comparableProvisional lane · 0 vs 2 weighted rows	Not comparable
Multilingual	Not ranked	Not ranked	Not comparableProvisional lane · 0 vs 0 weighted rows	Not comparable
Math	Not ranked	78.9Unranked · 2 rankable rows	Not comparableProvisional lane · 1 vs 2 weighted rows	Not comparable

Ranks count the models scored in each category’s lane, so the agentic, coding, and knowledge denominators (BenchAlign v5.8) differ from the provisional-lane categories. Unranked scores sit on the provisional lane’s scale but never name a winner. Like-for-like rows within 0.5 points read as a practical tie.

Supported evidence per lane · bars run 0–100Methodology

What each workload costs

Three fixed token mixes turn per-token rates into comparable decisions. Each scenario states context fit and whether cached input had to fall back to the published list-input rate.

Chat turn

1K fresh input + 500 output tokens

MiniMax M3: $0.0009; Fits in one request
Qwen3.8 Max: API rate not published; Fits in one request

Qwen3.8 Max has no comparable published API token rate.

Repository review

50K fresh input + 3K output tokens

MiniMax M3: $0.0186; Fits in one request
Qwen3.8 Max: API rate not published; Fits in one request

Qwen3.8 Max has no comparable published API token rate.

Cache-heavy agent loop

200K cached + 20K fresh input + 10K output tokens

MiniMax M3: $0.03; Fits in one request
Qwen3.8 Max: API rate not published; Fits in one request; Cached-input rate unavailable

Qwen3.8 Max has no comparable published API token rate.

Cached input falls back to the list input rate only where a cached rate is unpublished

Specification differences

Sourced differences are shown directly. Missing facts stay explicit instead of being inferred from a model name or family.

SpecificationMiniMax M3Qwen3.8 Max

Context window

Maximum documented context; output-token limits may be lower.

MiniMax M3

Qwen3.8 Max

Alibaba Cloud Model Studio pricing

API model ID

MiniMax M3

Not sourced

Qwen3.8 Max

qwen3.8-max

Alibaba Cloud Model Studio pricing

Cached-input rate

A missing cached-input rate falls back to the listed input rate only in the stated workload estimate.

MiniMax M3

$0.06 per 1M cached input tokens

Qwen3.8 Max

No comparable hosted API rate

Alibaba Cloud Model Studio pricing

Documented inputs

MiniMax M3

Not sourced

Qwen3.8 Max

Not sourced

Documented outputs

MiniMax M3

Not sourced

Qwen3.8 Max

Not sourced

Provider availability

MiniMax M3

Not sourced

Qwen3.8 Max

Not sourced

Reasoning profile

MiniMax M3

Non-Reasoning

Qwen3.8 Max

Reasoning

Weight access

MiniMax M3

Open Weight

Qwen3.8 Max

Open Weight

License

MiniMax M3

Open Weight

Qwen3.8 Max

Open Weight

Release date

MiniMax M3

2026-06-01

Qwen3.8 Max

2026-08-03

If you already use one of these models

Deployment change: The models list different providers, so authentication, endpoint behavior, limits, and feature support may change.
Quality signal: Qwen3.8 Max has the higher public score estimate, 64.86 versus 48.57, but the 90% score intervals overlap.
Workload cost: A complete comparable API-rate estimate is not available for both models.
Context tradeoff: Both models list 1M.

Run the same representative tasks against both endpoints before changing production traffic.

Questions

Which is better, MiniMax M3 or Qwen3.8 Max?

Qwen3.8 Max has the higher public score estimate, 64.86 versus 48.57, but the 90% score intervals overlap. The higher estimate is not a decisive winner because the uncertainty ranges overlap.

Which is better for coding, MiniMax M3 or Qwen3.8 Max?

Qwen3.8 Max has the higher public coding point estimate, 52.7 to 37.4, with Supported evidence for both models. The conditional ranges do not establish rank confidence.

Which is better for agentic tasks, MiniMax M3 or Qwen3.8 Max?

Qwen3.8 Max has the higher public agentic tasks point estimate, 66 to 39.3, with Supported evidence for both models. The conditional ranges do not establish rank confidence.

Which costs less, MiniMax M3 or Qwen3.8 Max?

Both models do not have comparable published API token rates, so this page does not name a universal price winner.

Which has the larger context window, MiniMax M3 or Qwen3.8 Max?

Both models list the same context window, 1M.

Self-host vs API cost

Estimates at 50,000 req/day · 1000 tokens/req average.

MiniMax M3

API / mo$1,125

Self-host / mo$13,981

Break-even818M/day

Qwen3.8 Max

API / mo$0

Self-host / mo$57,606

Break-even—

Model the full break-even

Benchmark evidence

The full public result ledger is available for audit without forcing a wide desktop table onto a phone.

Browse raw public benchmark evidence77 rows

Agentic

Terminal-Bench 2.1
MiniMax M366.0%
Source
Qwen3.8 Max86.6%
Source
Qwen3.8 Max leads this result
BrowseComp
MiniMax M383.5%
Source
Qwen3.8 Max—
Not directly comparable
OSWorld-Verified
MiniMax M370.1%
Source
Qwen3.8 Max86.1%
Source
Qwen3.8 Max leads this result
MCP Atlas
MiniMax M374.2%
Source
Qwen3.8 Max—
Not directly comparable
Claw-Eval
MiniMax M374.5%
Source
Qwen3.8 Max—
Not directly comparable
BankerToolBench
MiniMax M376.1%
Source
Qwen3.8 Max—
Not directly comparable
ResearchClawBench
MiniMax M319.8%
Source
Qwen3.8 Max—
Not directly comparable
OSWorld 2.0
MiniMax M34.6%
Source
Qwen3.8 Max19.4%
Source
Qwen3.8 Max leads this result
Terminal-Bench 2.1 (Vals)
MiniMax M353.6%
Source
Qwen3.8 Max67.4%
Source
Qwen3.8 Max leads this result
MazeBench
MiniMax M30%
Source
Qwen3.8 Max0%
Source
Tie
CoWorkBench
MiniMax M3—
Qwen3.8 Max74.8%
Source
Not directly comparable
JobBench
MiniMax M3—
Qwen3.8 Max53.4%
Source
Not directly comparable
skillsBench
MiniMax M3—
Qwen3.8 Max70.2%
Source
Not directly comparable
Agents' Last Exam
MiniMax M3—
Qwen3.8 Max52.4%
Source
Not directly comparable
AutomationBench
MiniMax M3—
Qwen3.8 Max27.3%
Source
Not directly comparable
Toolathlon-Verified
MiniMax M3—
Qwen3.8 Max72.5%
Source
Not directly comparable
WideResearch
MiniMax M3—
Qwen3.8 Max81.9%
Source
Not directly comparable
HLE w/ tools
MiniMax M3—
Qwen3.8 Max56.2%
Source
Not directly comparable
WebArena-Verified
MiniMax M3—
Qwen3.8 Max66.8%
Source
Not directly comparable
AndroidWorld
MiniMax M3—
Qwen3.8 Max85.3%
Source
Not directly comparable
MobileWorld
MiniMax M3—
Qwen3.8 Max77.8%
Source
Not directly comparable
Hermes Index
MiniMax M3—
Qwen3.8 Max36.2%
Source
Not directly comparable
Hermes Bench
MiniMax M3—
Qwen3.8 Max70.5%
Source
Not directly comparable

Coding

SWE-bench Verified
MiniMax M380.5%
Source
Qwen3.8 Max—
Not directly comparable
SWE-bench Pro
MiniMax M359%
Source
Qwen3.8 Max67.7%
Source
Qwen3.8 Max leads this result
Terminal-Bench 2.1
MiniMax M366.0%
Source
Qwen3.8 Max86.6%
Source
Qwen3.8 Max leads this result
NL2Repo
MiniMax M342.1%
Source
Qwen3.8 Max55.9%
Source
Qwen3.8 Max leads this result
VIBE V2
MiniMax M350.1%
Source
Qwen3.8 Max—
Not directly comparable
SVG-Bench
MiniMax M363.7%
Source
Qwen3.8 Max—
Not directly comparable
KernelBench Hard
MiniMax M328.8%
Source
Qwen3.8 Max—
Not directly comparable
OpenHarmony Bench
Shared source
MiniMax M348.4%
Qwen3.8 Max60.8%
Qwen3.8 Max leads this result
LiveCodeBench (Vals)
MiniMax M382.2%
Source
Qwen3.8 Max87.9%
Source
Qwen3.8 Max leads this result
SWE-bench (Vals)
MiniMax M375.0%
Source
Qwen3.8 Max85.6%
Source
Qwen3.8 Max leads this result
DeepSWE
MiniMax M3—
Qwen3.8 Max56.6%
Source
Not directly comparable
FrontierSWE
MiniMax M3—
Qwen3.8 Max73.5%
Source
Not directly comparable
MLS-Bench Lite
MiniMax M3—
Qwen3.8 Max41.0%
Source
Not directly comparable
PaperBench
MiniMax M3—
Qwen3.8 Max93.0%
Source
Not directly comparable
VulcanBench v3
MiniMax M3—
Qwen3.8 Max81.2%
Source
Not directly comparable
FrontierSWE v2
MiniMax M3—
Qwen3.8 Max15.8%
Source
Not directly comparable
Bug Hunt Bench
MiniMax M3—
Qwen3.8 Max25.7 fixes
Source
Not directly comparable

Reasoning

MRCRv2
MiniMax M3—
Qwen3.8 Max92.9%
Source
Not directly comparable
LongBench v2
MiniMax M3—
Qwen3.8 Max66.3%
Source
Not directly comparable

Multimodal

OfficeQA Pro
MiniMax M345.1%
Source
Qwen3.8 Max—
Not directly comparable
OmniDocBench 1.5
MiniMax M391.6%
Source
Qwen3.8 Max92.1%
Source
Qwen3.8 Max leads this result
MMMU-Pro
MiniMax M378.1%
Source
Qwen3.8 Max82.3%
Source
Qwen3.8 Max leads this result
VideoMMMU
MiniMax M384.6%
Source
Qwen3.8 Max88.7%
Source
Qwen3.8 Max leads this result
Video-MME (with subtitle)
MiniMax M385.4%
Source
Qwen3.8 Max90.4%
Source
Qwen3.8 Max leads this result
MathVision
MiniMax M3—
Qwen3.8 Max95.2%
Source
Not directly comparable
MathVision w/ Python
MiniMax M3—
Qwen3.8 Max97.7%
Source
Not directly comparable
BabyVision
MiniMax M3—
Qwen3.8 Max82.0%
Source
Not directly comparable
BabyVision w/ Python
MiniMax M3—
Qwen3.8 Max91.3%
Source
Not directly comparable
ZeroBench
MiniMax M3—
Qwen3.8 Max24.0%
Source
Not directly comparable
ZeroBench w/ Python
MiniMax M3—
Qwen3.8 Max49.0%
Source
Not directly comparable
MedXpertQA (MM)
MiniMax M3—
Qwen3.8 Max80.4%
Source
Not directly comparable
ScreenSpot Pro
MiniMax M3—
Qwen3.8 Max84.5%
Source
Not directly comparable
Vision2Web
MiniMax M3—
Qwen3.8 Max69.0%
Source
Not directly comparable
CharXiv w/o tools
MiniMax M3—
Qwen3.8 Max88.4%
Source
Not directly comparable
CharXiv
MiniMax M3—
Qwen3.8 Max93.5%
Source
Not directly comparable
OCRBench V2
MiniMax M3—
Qwen3.8 Max74.2%
Source
Not directly comparable
CC-OCR
MiniMax M3—
Qwen3.8 Max79.6%
Source
Not directly comparable
RealWorldQA
MiniMax M3—
Qwen3.8 Max88.0%
Source
Not directly comparable
ERQA
MiniMax M3—
Qwen3.8 Max77.8%
Source
Not directly comparable
SimpleVQA
MiniMax M3—
Qwen3.8 Max75.0%
Source
Not directly comparable
PerceptionBench
MiniMax M3—
Qwen3.8 Max63.5%
Source
Not directly comparable
MMVU
MiniMax M3—
Qwen3.8 Max82.4%
Source
Not directly comparable
MLVU (M-Avg)
MiniMax M3—
Qwen3.8 Max90.8%
Source
Not directly comparable
LVBench
MiniMax M3—
Qwen3.8 Max81.8%
Source
Not directly comparable

Knowledge

GPQA Diamond (Vals)
MiniMax M392.7%
Source
Qwen3.8 Max93.7%
Source
Qwen3.8 Max leads this result
MMLU-Pro (Vals)
MiniMax M384.2%
Source
Qwen3.8 Max88.6%
Source
Qwen3.8 Max leads this result
GPQA
MiniMax M3—
Qwen3.8 Max92.6%
Source
Not directly comparable
GPQA-D
MiniMax M3—
Qwen3.8 Max92.6%
Source
Not directly comparable
HLE
MiniMax M3—
Qwen3.8 Max43.6%
Source
Not directly comparable
HLE w/o tools
MiniMax M3—
Qwen3.8 Max43.6%
Source
Not directly comparable

Instruction following

IFBench
MiniMax M3—
Qwen3.8 Max82.8%
Source
Not directly comparable

Math

USAMO 2026
MiniMax M385.7%
Source
Qwen3.8 Max—
Not directly comparable
FrontierMath v2 (Tiers 1-3)
MiniMax M3—
Qwen3.8 Max74.737%
Source
Not directly comparable
FrontierMath v2 (Tier 4)
MiniMax M3—
Qwen3.8 Max46.341%
Source
Not directly comparable

77 public results · 17 shared

Related comparisons

Compare API pricing Read the methodology Open the model selector

Watch MiniMax M3 vs Qwen3.8 Max

One weekly email when material rank, price, or benchmark evidence changes make this matchup worth revisiting.

Read a sample issue

Join 6,100+ subscribers.

Last updated October 29, 2026

MiniMax M3 vs Qwen3.8 Max

Decision reading

Which one for your work

Coding work

Agentic work

Long documents

Chat turn cost

Cache-heavy agent loop cost

Repository review cost

Which one for a specific job

What is actually comparable

Shape of the matched evidence

Category results, on a stated basis

Agentic

Coding

Knowledge

Multimodal

Instruction following

Reasoning

Multilingual

Math

What each workload costs

Chat turn

Repository review

Cache-heavy agent loop

Specification differences

Context window

API model ID

Cached-input rate

Documented inputs

Documented outputs

Provider availability

Reasoning profile

Weight access

License

Release date

If you already use one of these models

Questions

Which is better, MiniMax M3 or Qwen3.8 Max?

Which is better for coding, MiniMax M3 or Qwen3.8 Max?

Which is better for agentic tasks, MiniMax M3 or Qwen3.8 Max?

Which costs less, MiniMax M3 or Qwen3.8 Max?

Which has the larger context window, MiniMax M3 or Qwen3.8 Max?

Self-host vs API cost

Benchmark evidence

Agentic

Coding

Reasoning

Multimodal

Knowledge

Instruction following

Math

Related comparisons

Watch MiniMax M3 vs Qwen3.8 Max