Coding work
Code generation, repair, and software-engineering tasks
Qwen3.8-27B
Qwen3.8-27B leads on the public coding lane, 48.7 to 38.7, with Supported evidence for both models, although the 90% intervals overlap.
Confidence: limited
MiniMax M3 vs Qwen3.8-27B
Updated September 28, 2026. Rank cannot separate these two. Price, access, and your workload decide. Public scores include evidence status and uncertainty.
Both of these models will change. Get the price, version and retirement notices for the pair, each with its source. Follow model changes
Share or export
Decision reading
Qwen3.8-27B has the higher public score estimate, 55.26 versus 54.78, but the 90% score intervals overlap. Treat that as a lead, not a settled winner. 11 results are shared. Category rows resting on Estimated evidence or different benchmark sets are marked directional and do not name a winner.
- Shared results
- 11
- MiniMax M3 only
- 16
- Qwen3.8-27B only
- 22
- Like-for-like categories
- 3 / 8
Supported: MiniMax M3 · Estimated: Qwen3.8-27BHow 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.
Agentic work
Tool use, computer use, and multi-step task completion
Qwen3.8-27B
Qwen3.8-27B leads on the public agentic lane, 61.2 to 39.7, with Supported evidence for both models and non-overlapping 90% intervals.
Confidence: strongerLong documents
Prompts that approach the documented context limit
MiniMax M3
MiniMax M3 has the larger documented context window.
Confidence: documented
Show secondary and unsupported calls
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-ratesCache-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-ratesRepository 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.
The same task catalog as the LLM Selector. Each job names the evidence surface it rests on; nothing here adds a new score.
38.7MiniMax M348.7Qwen3.8-27B
Like-for-like · BenchAlign v5.7
Qwen3.8-27B leads the like-for-like coding row, although the 90% intervals overlap.
Coding scores combine specific tasks and setups. Match the editor, harness, and effort to your workflow.
Same basis rules as the category table below
What is actually comparable
Shared results can support a head-to-head reading. Results present for only one model describe coverage, not superiority.
16 MiniMax M3 only
11 Shared
22 Qwen3.8-27B only
3 categories rest on Estimated evidence or different evidence sets. Those rows remain visible for coverage context but do not name a winner.
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 14.2MiniMax M3:70.1%Qwen3.8-27B:84.3%SWE-bench ProCoding
Normalized gap 2.7MiniMax M3:59%Qwen3.8-27B:61.7%LiveCodeBench (Vals)Coding
Normalized gap 1.8MiniMax M3:82.2%Qwen3.8-27B:84.0%MMLU-Pro (Vals)Knowledge
Normalized gap 0.1MiniMax M3:84.2%Qwen3.8-27B:84.3%
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 v5.7 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
- 39.7
- Supported · #52/111
- Qwen3.8-27B
- 61.2
- Supported · #16/111
- Basis
- BenchAlign v5.7 lane · 9 vs 8 public rows
- Reading
- Qwen3.8-27B leads
Coding
Like-for-like- MiniMax M3
- 38.7
- Supported · #59/136
- Qwen3.8-27B
- 48.7
- Supported · #41/136
- Basis
- BenchAlign v5.7 lane · 10 vs 8 public rows
- Reading
- Qwen3.8-27B leads · intervals overlap
Knowledge
Like-for-like- MiniMax M3
- 48.3
- Supported · #63/160
- Qwen3.8-27B
- 49.2
- Supported · #59/160
- Basis
- BenchAlign v5.7 lane · 2 vs 6 public rows
- Reading
- Qwen3.8-27B leads · intervals overlap
Reasoning
Directional only- MiniMax M3
- 79.4
- #5/27
- Qwen3.8-27B
- 78.7
- #8/27
- Basis
- Provisional lane · 0 vs 0 weighted rows
- Reading
- Directional only
Multimodal
Directional only- MiniMax M3
- 52.0
- #37/50
- Qwen3.8-27B
- 80.9
- #11/50
- Basis
- Provisional lane · 2 vs 1 weighted rows
- Reading
- Directional only
Instruction following
Directional only- MiniMax M3
- 92.4
- #5/124
- Qwen3.8-27B
- 83.2
- #45/124
- Basis
- Provisional lane · 0 vs 1 weighted rows
- Reading
- Directional only
Multilingual
Not comparable- MiniMax M3
- Not ranked
- Qwen3.8-27B
- Not ranked
- Basis
- Provisional lane · 0 vs 0 weighted rows
- Reading
- Not comparable
Math
Not comparable- MiniMax M3
- Not ranked
- Qwen3.8-27B
- Not ranked
- Basis
- Provisional lane · 1 vs 0 weighted rows
- Reading
- Not comparable
| Category | MiniMax M3 | Qwen3.8-27B | Basis | Reading |
|---|---|---|---|---|
| Agentic | 39.7Supported · #52/111 | 61.2Supported · #16/111 | Like-for-likeBenchAlign v5.7 lane · 9 vs 8 public rows | Qwen3.8-27B leads |
| Coding | 38.7Supported · #59/136 | 48.7Supported · #41/136 | Like-for-likeBenchAlign v5.7 lane · 10 vs 8 public rows | Qwen3.8-27B leads · intervals overlap |
| Knowledge | 48.3Supported · #63/160 | 49.2Supported · #59/160 | Like-for-likeBenchAlign v5.7 lane · 2 vs 6 public rows | Qwen3.8-27B leads · intervals overlap |
| Reasoning | 79.4#5/27 | 78.7#8/27 | Directional onlyProvisional lane · 0 vs 0 weighted rows | Directional only |
| Multimodal | 52.0#37/50 | 80.9#11/50 | Directional onlyProvisional lane · 2 vs 1 weighted rows | Directional only |
| Instruction following | 92.4#5/124 | 83.2#45/124 | Directional onlyProvisional lane · 0 vs 1 weighted rows | Directional only |
| Multilingual | Not ranked | Not ranked | Not comparableProvisional lane · 0 vs 0 weighted rows | Not comparable |
| Math | Not ranked | Not ranked | Not comparableProvisional lane · 1 vs 0 weighted rows | Not comparable |
Ranks count the models scored in each category’s lane, so the agentic, coding, and knowledge denominators (BenchAlign v5.7) 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
- Qwen3.8-27B
- Self-hosted; infrastructure cost varies
- Fits in one request
Qwen3.8-27B has no comparable published API token rate.
Repository review
50K fresh input + 3K output tokens
- MiniMax M3
- $0.0186
- Fits in one request
- Qwen3.8-27B
- Self-hosted; infrastructure cost varies
- Fits in one request
Qwen3.8-27B 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
- Qwen3.8-27B
- Self-hosted; infrastructure cost varies
- Fits in one request
- Cached-input rate unavailable
Qwen3.8-27B 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 M3Qwen3.8-27B
Context window
Maximum documented context; output-token limits may be lower.
MiniMax M3
1M
Qwen3.8-27B
API model ID
MiniMax M3
Not sourced
Qwen3.8-27B
Not sourced
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
Qwen3.8-27B
No comparable hosted API rate
Qwen3.8-27B model cardDocumented inputs
MiniMax M3
Not sourced
Qwen3.8-27B
Not sourced
Documented outputs
MiniMax M3
Not sourced
Qwen3.8-27B
Not sourced
Provider availability
MiniMax M3
Not sourced
Qwen3.8-27B
Not sourced
Reasoning profile
MiniMax M3
Non-Reasoning
Qwen3.8-27B
Reasoning
Weight access
MiniMax M3
Open Weight
Qwen3.8-27B
Open Weight
License
MiniMax M3
Open Weight
Qwen3.8-27B
Open Weight
Release date
MiniMax M3
2026-06-01
Qwen3.8-27B
2026-08-05
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
- Qwen3.8-27B has the higher public score estimate, 55.26 versus 54.78, but the 90% score intervals overlap.
- Workload cost
- A complete comparable API-rate estimate is not available for both models.
- Context tradeoff
- MiniMax M3 has the larger documented window (1M).
Run the same representative tasks against both endpoints before changing production traffic.
Questions
Which is better, MiniMax M3 or Qwen3.8-27B?
Qwen3.8-27B has the higher public score estimate, 55.26 versus 54.78, 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 Qwen3.8-27B?
Qwen3.8-27B leads the public coding lane, 48.7 to 38.7, with Supported evidence for both models, although the 90% intervals overlap.
Which is better for agentic tasks, MiniMax M3 or Qwen3.8-27B?
Qwen3.8-27B leads the public agentic tasks lane, 61.2 to 39.7, with Supported evidence for both models and non-overlapping 90% intervals.
Which costs less, MiniMax M3 or Qwen3.8-27B?
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 Qwen3.8-27B?
MiniMax M3 has the larger documented context window: 1M, compared with 262K.
Benchmark evidence
The full public result ledger is available for audit without forcing a wide desktop table onto a phone.
Browse raw public benchmark evidence49 rows
Agentic
Terminal-Bench 2.1
MiniMax M366.0%SourceQwen3.8-27B73.0%SourceQwen3.8-27B leads this result
BrowseComp
MiniMax M383.5%SourceQwen3.8-27B—Not directly comparable
OSWorld-Verified
MiniMax M370.1%SourceQwen3.8-27B84.3%SourceQwen3.8-27B leads this result
MCP Atlas
MiniMax M374.2%SourceQwen3.8-27B—Not directly comparable
Claw-Eval
MiniMax M374.5%SourceQwen3.8-27B—Not directly comparable
BankerToolBench
MiniMax M376.1%SourceQwen3.8-27B—Not directly comparable
ResearchClawBench
MiniMax M319.8%SourceQwen3.8-27B—Not directly comparable
OSWorld 2.0
MiniMax M34.6%SourceQwen3.8-27B—Not directly comparable
Terminal-Bench 2.1 (Vals)
MiniMax M353.6%SourceQwen3.8-27B58.4%SourceQwen3.8-27B leads this result
CoWorkBench
MiniMax M3—Qwen3.8-27B70.7%SourceNot directly comparable
JobBench
MiniMax M3—Qwen3.8-27B33.4%SourceNot directly comparable
Agents' Last Exam
MiniMax M3—Qwen3.8-27B42.9%SourceNot directly comparable
WebArena-Verified
MiniMax M3—Qwen3.8-27B64.8%SourceNot directly comparable
AndroidWorld
MiniMax M3—Qwen3.8-27B81.9%SourceNot directly comparable
Coding
SWE-bench Verified
MiniMax M380.5%SourceQwen3.8-27B—Not directly comparable
SWE-bench Pro
MiniMax M359%SourceQwen3.8-27B61.7%SourceQwen3.8-27B leads this result
Terminal-Bench 2.1
MiniMax M366.0%SourceQwen3.8-27B73.0%SourceQwen3.8-27B leads this result
NL2Repo
MiniMax M342.1%SourceQwen3.8-27B42.3%SourceQwen3.8-27B leads this result
VIBE V2
MiniMax M350.1%SourceQwen3.8-27B—Not directly comparable
SVG-Bench
MiniMax M363.7%SourceQwen3.8-27B—Not directly comparable
KernelBench Hard
MiniMax M328.8%SourceQwen3.8-27B—Not directly comparable
OpenHarmony Bench
MiniMax M348.4%SourceQwen3.8-27B—Not directly comparable
LiveCodeBench (Vals)
MiniMax M382.2%SourceQwen3.8-27B84.0%SourceQwen3.8-27B leads this result
SWE-bench (Vals)
MiniMax M375.0%SourceQwen3.8-27B86.0%SourceQwen3.8-27B leads this result
DeepSWE
MiniMax M3—Qwen3.8-27B42.2%SourceNot directly comparable
LiveCodeBench v6
MiniMax M3—Qwen3.8-27B90.3%SourceNot directly comparable
VulcanBench v3
MiniMax M3—Qwen3.8-27B82.6%SourceNot directly comparable
Multimodal
OfficeQA Pro
MiniMax M345.1%SourceQwen3.8-27B—Not directly comparable
OmniDocBench 1.5
MiniMax M391.6%SourceQwen3.8-27B91.1%SourceMiniMax M3 leads this result
MMMU-Pro
MiniMax M378.1%SourceQwen3.8-27B—Not directly comparable
VideoMMMU
MiniMax M384.6%SourceQwen3.8-27B—Not directly comparable
Video-MME (with subtitle)
MiniMax M385.4%SourceQwen3.8-27B—Not directly comparable
MathVision
MiniMax M3—Qwen3.8-27B90.0%SourceNot directly comparable
MathVision w/ Python
MiniMax M3—Qwen3.8-27B94.6%SourceNot directly comparable
BabyVision
MiniMax M3—Qwen3.8-27B65.7%SourceNot directly comparable
BabyVision w/ Python
MiniMax M3—Qwen3.8-27B85.6%SourceNot directly comparable
Vision2Web
MiniMax M3—Qwen3.8-27B62.9%SourceNot directly comparable
CharXiv w/o tools
MiniMax M3—Qwen3.8-27B83.7%SourceNot directly comparable
CharXiv
MiniMax M3—Qwen3.8-27B90.2%SourceNot directly comparable
RealWorldQA
MiniMax M3—Qwen3.8-27B85.9%SourceNot directly comparable
ERQA
MiniMax M3—Qwen3.8-27B65.5%SourceNot directly comparable
Knowledge
GPQA Diamond (Vals)
MiniMax M392.7%SourceQwen3.8-27B88.9%SourceMiniMax M3 leads this result
MMLU-Pro (Vals)
MiniMax M384.2%SourceQwen3.8-27B84.3%SourceQwen3.8-27B leads this result
GPQA
MiniMax M3—Qwen3.8-27B89.2%SourceNot directly comparable
GPQA-D
MiniMax M3—Qwen3.8-27B89.2%SourceNot directly comparable
HLE
MiniMax M3—Qwen3.8-27B30.8%SourceNot directly comparable
HLE w/o tools
MiniMax M3—Qwen3.8-27B30.8%SourceNot directly comparable
Instruction following
IFBench
MiniMax M3—Qwen3.8-27B79.5%SourceNot directly comparable
Math
USAMO 2026
MiniMax M385.7%SourceQwen3.8-27B—Not directly comparable
49 public results · 11 shared
Watch MiniMax M3 vs Qwen3.8-27B
One weekly email when material rank, price, or benchmark evidence changes make this matchup worth revisiting.
Read a sample issueJoin 2,000+ readers.
Last updated September 28, 2026