Agentic
Like-for-like- MiniMax M3
- 72.3
- Qwen3.5-27B
- 52.0
- Weighted basis
- 3 vs 3 rows
- Reading
- MiniMax M3 leads
MiniMax M3 vs Qwen3.5-27B
Updated August 30, 2026. Public scores include evidence status and uncertainty. They are not guarantees for a specific workload.
Decision reading
MiniMax M3 has the higher public score estimate, 68.73 versus 60.04, but the 90% score intervals overlap. Treat that as a lead, not a settled winner.
4 results are shared. Category rows based on different benchmark sets are marked directional and do not name a winner.
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
MiniMax M3
MiniMax M3 leads on the same 3 weighted benchmark rows.
Confidence: stronger
Long 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
Coding work
Code generation, repair, and software-engineering tasks
Not enough matched evidence
The category averages use different weighted benchmark sets, so they are directional rather than like-for-like.
Confidence: limited
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
What is actually comparable
Shared results can support a head-to-head reading. Results present for only one model describe coverage, not superiority.
18 MiniMax M3 only
4 Shared
12 Qwen3.5-27B only
- Shared results
- 4
- MiniMax M3 only
- 18
- Qwen3.5-27B only
- 12
- Like-for-like categories
- 1 / 8
1 category uses different evidence sets. Those rows remain visible for coverage context but do not name a winner.
Category results, on a stated basis
Each row states whether both averages use the same weighted benchmark set. Directional and not-comparable rows remain visible, but they never receive a winner in this template.
Coding
Directional only- MiniMax M3
- 72.2
- Qwen3.5-27B
- 64.9
- Weighted basis
- 2 vs 2 rows
- Reading
- Directional only
Reasoning
Not comparable- MiniMax M3
- Not measured
- Qwen3.5-27B
- 60.6
- Weighted basis
- 0 vs 1 rows
- Reading
- Not comparable
Knowledge
Not comparable- MiniMax M3
- Not measured
- Qwen3.5-27B
- 82.7
- Weighted basis
- 0 vs 3 rows
- Reading
- Not comparable
Math
Not comparable- MiniMax M3
- 85.7
- Qwen3.5-27B
- Not measured
- Weighted basis
- 1 vs 0 rows
- Reading
- Not comparable
Multilingual
Not comparable- MiniMax M3
- Not measured
- Qwen3.5-27B
- 82.2
- Weighted basis
- 0 vs 1 rows
- Reading
- Not comparable
Multimodal
Not comparable- MiniMax M3
- 64.9
- Qwen3.5-27B
- Not measured
- Weighted basis
- 2 vs 0 rows
- Reading
- Not comparable
Instruction following
Not comparable- MiniMax M3
- Not measured
- Qwen3.5-27B
- 95.0
- Weighted basis
- 0 vs 1 rows
- Reading
- Not comparable
| Category | MiniMax M3 | Qwen3.5-27B | Weighted basis | Reading |
|---|---|---|---|---|
| Agentic | 72.3 | 52.0 | Like-for-like3 vs 3 rows | MiniMax M3 leads |
| Coding | 72.2 | 64.9 | Directional only2 vs 2 rows | Directional only |
| Reasoning | Not measured | 60.6 | Not comparable0 vs 1 rows | Not comparable |
| Knowledge | Not measured | 82.7 | Not comparable0 vs 3 rows | Not comparable |
| Math | 85.7 | Not measured | Not comparable1 vs 0 rows | Not comparable |
| Multilingual | Not measured | 82.2 | Not comparable0 vs 1 rows | Not comparable |
| Multimodal | 64.9 | Not measured | Not comparable2 vs 0 rows | Not comparable |
| Instruction following | Not measured | 95.0 | Not comparable0 vs 1 rows | Not comparable |
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.
Terminal-Bench 2.0
Agentic
BrowseComp
Agentic
OSWorld-Verified
Agentic
SWE-bench Verified
Coding
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
- Qwen3.5-27B
- Self-hosted; infrastructure cost varies
- Fits in one request
Qwen3.5-27B has no comparable published API token rate.
Repository review
50K fresh input + 3K output tokens
- MiniMax M3
- $0.0186
- Fits in one request
- Qwen3.5-27B
- Self-hosted; infrastructure cost varies
- Fits in one request
Qwen3.5-27B 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
- Qwen3.5-27B
- Self-hosted; infrastructure cost varies
- Fits in one request
- Cached-input rate unavailable
Qwen3.5-27B has no comparable published API token rate.
Specification differences
Sourced differences are shown directly. Missing facts stay explicit instead of being inferred from a model name or family.
SpecificationMiniMax M3Qwen3.5-27B
Context window
Maximum documented context; output-token limits may be lower.
MiniMax M3
1M
Qwen3.5-27B
262K
API model ID
MiniMax M3
Not sourced
Qwen3.5-27B
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
Qwen3.5-27B
No comparable hosted API rate
Documented inputs
MiniMax M3
Not sourced
Qwen3.5-27B
Not sourced
Documented outputs
MiniMax M3
Not sourced
Qwen3.5-27B
Not sourced
Provider availability
MiniMax M3
Not sourced
Qwen3.5-27B
Not sourced
Reasoning profile
MiniMax M3
Non-Reasoning
Qwen3.5-27B
Reasoning
Weight access
MiniMax M3
Open Weight
Qwen3.5-27B
Open Weight
License
MiniMax M3
Open Weight
Qwen3.5-27B
Open Weight
Release date
MiniMax M3
2026-06-01
Qwen3.5-27B
2026-03-04
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
- MiniMax M3 has the higher public score estimate, 68.73 versus 60.04, 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.
Benchmark evidence
The full public result ledger is available for audit without forcing a wide desktop table onto a phone.
Browse raw public benchmark evidence34 rows
Agentic
Terminal-Bench 2.0
MiniMax M366%SourceQwen3.5-27B41.6%SourceMiniMax M3 leads this result
BrowseComp
MiniMax M383.5%SourceQwen3.5-27B61%SourceMiniMax M3 leads this result
OSWorld-Verified
MiniMax M370.1%SourceQwen3.5-27B56.2%SourceMiniMax M3 leads this result
MCP Atlas
MiniMax M374.2%SourceQwen3.5-27B—Not directly comparable
Claw-Eval
MiniMax M374.5%SourceQwen3.5-27B—Not directly comparable
BankerToolBench
MiniMax M376.1%SourceQwen3.5-27B—Not directly comparable
ResearchClawBench
MiniMax M319.8%SourceQwen3.5-27B—Not directly comparable
OSWorld 2.0
MiniMax M34.6%SourceQwen3.5-27B—Not directly comparable
Gert Labs
MiniMax M3—Qwen3.5-27B39.41%SourceNot directly comparable
Coding
SWE-bench Verified
MiniMax M380.5%SourceQwen3.5-27B72.4%SourceMiniMax M3 leads this result
SWE-bench Pro
MiniMax M359%SourceQwen3.5-27B—Not directly comparable
Terminal-Bench 2.0
MiniMax M366.0%SourceQwen3.5-27B—Not directly comparable
NL2Repo
MiniMax M342.1%SourceQwen3.5-27B—Not directly comparable
VIBE V2
MiniMax M350.1%SourceQwen3.5-27B—Not directly comparable
SVG-Bench
MiniMax M363.7%SourceQwen3.5-27B—Not directly comparable
KernelBench Hard
MiniMax M328.8%SourceQwen3.5-27B—Not directly comparable
OpenHarmony Bench
MiniMax M348.4%SourceQwen3.5-27B—Not directly comparable
SWE-Rebench
MiniMax M3—Qwen3.5-27B58.9%SourceNot directly comparable
Reasoning
LongBench v2
MiniMax M3—Qwen3.5-27B60.6%SourceNot directly comparable
Knowledge
Math
USAMO 2026
MiniMax M385.7%SourceQwen3.5-27B—Not directly comparable
Multilingual
MMLU-ProX
MiniMax M3—Qwen3.5-27B82.2%SourceNot directly comparable
Multimodal
OfficeQA Pro
MiniMax M345.1%SourceQwen3.5-27B—Not directly comparable
OmniDocBench 1.5
MiniMax M391.6%SourceQwen3.5-27B—Not directly comparable
MMMU-Pro
MiniMax M378.1%SourceQwen3.5-27B—Not directly comparable
VideoMMMU
MiniMax M384.6%SourceQwen3.5-27B—Not directly comparable
Video-MME (with subtitle)
MiniMax M385.4%SourceQwen3.5-27B—Not directly comparable
MMMU
MiniMax M3—Qwen3.5-27B82.3%SourceNot directly comparable
MMVU
MiniMax M3—Qwen3.5-27B73.3%SourceNot directly comparable
MathVision
MiniMax M3—Qwen3.5-27B86.0%SourceNot directly comparable
V*
MiniMax M3—Qwen3.5-27B93.7%SourceNot directly comparable
Instruction following
IFEval
MiniMax M3—Qwen3.5-27B95%SourceNot directly comparable
Frequently asked questions
Which is better, MiniMax M3 or Qwen3.5-27B?
MiniMax M3 has the higher public score estimate, 68.73 versus 60.04, 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 Qwen3.5-27B?
The current coding averages use different weighted benchmark sets, so BenchLM does not name a winner from them. Read the shared benchmark rows directly and test the models on the same task set.
Which is better for agentic tasks, MiniMax M3 or Qwen3.5-27B?
MiniMax M3 leads the like-for-like agentic tasks comparison across 3 shared weighted benchmark rows.
Which costs less, MiniMax M3 or Qwen3.5-27B?
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 Qwen3.5-27B?
MiniMax M3 has the larger documented context window: 1M, compared with 262K.
Related comparisons
Last updated August 30, 2026
Watch MiniMax M3 vs Qwen3.5-27B
One weekly email when material rank, price, or benchmark evidence changes make this matchup worth revisiting.
Read a sample issueJoin 2,000+ readers.