Skip to main content

Model comparison

Mellum2-12B-A2.5B-Thinking vs Qwen3.5 397B

Data verified

Head-to-head evidence from 3 shared benchmark results across 2 categories. Overall scores shown here use the public BenchAlign v5 ranking lane.

No comparison
57.01/100
1 category wins1 category wins

Public leaderboard positions: Mellum2-12B-A2.5B-Thinking unranked (Not scored); Qwen3.5 397B #71 (Estimated). Intervals and evidence labels describe ranking uncertainty, not a guarantee for a specific workload.

Evidence parity. Mellum2-12B-A2.5B-Thinking and Qwen3.5 397B share 3 comparable benchmark results. 2 of 8 categories are comparable. 2 results are unique to Mellum2-12B-A2.5B-Thinking; 52 to Qwen3.5 397B.

Updated July 23, 2026
Shared results
3
Mellum2-12B-A2.5B-Thinking only
2
Qwen3.5 397B only
52
Comparable categories
2 / 8

Treat this as a split decision. Mellum2-12B-A2.5B-Thinking makes more sense if knowledge is the priority or you want the stronger reasoning-first profile; Qwen3.5 397B is the better fit if instruction following is the priority or you would rather avoid the extra latency and token burn of a reasoning model.

Confidence note. This is a partial-evidence comparison with 3 shared benchmark results across 2 evidence categories; 2 of 8 categories currently have scoreable aggregates for both models. Treat the verdict as directional until coverage is more balanced.

Why this result

Mellum2-12B-A2.5B-Thinking and Qwen3.5 397B finish on the same BenchAlign overall score, so this is less about a single winner and more about where the edge shows up. The BenchAlign headline says tie; the benchmark table is where the real choice happens.

Mellum2-12B-A2.5B-Thinking is the reasoning model in the pair, while Qwen3.5 397B is not. That usually helps on harder chain-of-thought-heavy tests, but it can also mean more latency and more token spend in real use.

Category breakdown

Exact category averages are shown below. Not measured means BenchLM does not have enough sourced public coverage for that model and category.

Category scores and score margins for Mellum2-12B-A2.5B-Thinking and Qwen3.5 397B
CategoryMellum2-12B-A2.5B-ThinkingΔQwen3.5 397B
Inst. FollowingMellum2-12B-A2.5B-Thinking76.5Margin 16.1Qwen3.5 397B92.6
KnowledgeMellum2-12B-A2.5B-Thinking57.6Margin 1.0Qwen3.5 397B56.6
AgenticMellum2-12B-A2.5B-ThinkingNot measuredMarginNo overlapQwen3.5 397B56.5
CodingMellum2-12B-A2.5B-ThinkingNot measuredMarginNo overlapQwen3.5 397B66.5
ReasoningMellum2-12B-A2.5B-ThinkingNot measuredMarginNo overlapQwen3.5 397B63.2
MathMellum2-12B-A2.5B-ThinkingNot measuredMarginNo overlapQwen3.5 397B90.6
MultilingualMellum2-12B-A2.5B-ThinkingNot measuredMarginNo overlapQwen3.5 397B84.7
MultimodalMellum2-12B-A2.5B-ThinkingNot measuredMarginNo overlapQwen3.5 397B79.6

Decisive benchmark drivers

The largest measured benchmark gaps in this matchup, with exact reported values.

More
A · Mellum2-12B-A2.5B-ThinkingB · Qwen3.5 397B
  1. GPQA

    Knowledge
    Source ↗
    A 57.6%B 88.4%
    Winner: Qwen3.5 397BΔ 30.8
    GPQA: Mellum2-12B-A2.5B-Thinking scored 57.6%; Qwen3.5 397B scored 88.4%. Qwen3.5 397B wins this benchmark.
  2. IFEval

    Inst. Following
    Source ↗
    A 76.5%B 92.6%
    Winner: Qwen3.5 397BΔ 16.1
    IFEval: Mellum2-12B-A2.5B-Thinking scored 76.5%; Qwen3.5 397B scored 92.6%. Qwen3.5 397B wins this benchmark.

Operational comparison

Runtime and commercial metrics are compared only when both models have a complete sourced value.

MetricMellum2-12B-A2.5B-ThinkingQwen3.5 397BComparison
Input / output priceUSD per 1M tokensMellum2-12B-A2.5B-ThinkingNot availableQwen3.5 397B$0.6 input / $3.6 outputA complete price comparison is not available.
Generation speedtokens per secondMellum2-12B-A2.5B-ThinkingNot availableQwen3.5 397B96 tok/sA complete speed comparison is not available.
First-answer latencyseconds to first tokenMellum2-12B-A2.5B-ThinkingNot availableQwen3.5 397B2.44 sA complete latency comparison is not available.
Context windowmaximum listed tokensMellum2-12B-A2.5B-Thinking128KQwen3.5 397B128KListed context windows are equal.

Benchmark Deep Dive

Agentic
BenchmarkMellum2-12B-A2.5B-ThinkingQwen3.5 397BResult
BFCL v4Source 45.6%Not comparable
Terminal-Bench 2.0Source 52.5%Not comparable
BrowseCompSource 62%Not comparable
Claw-EvalSource 56.8%Not comparable
QwenClawBenchSource 51.8%Not comparable
τ³-bench resultsSource 68.4%Not comparable
VITA-BenchSource 43.7%Not comparable
DeepPlanningSource 37.6%Not comparable
ToolathlonSource 36.3%Not comparable
MCP AtlasSource 46.1%Not comparable
MCP-TasksSource 74.2%Not comparable
WideResearchSource 74.0%Not comparable
τ²-bench resultsSource 95.6%Not comparable
Gert LabsSource 46.76%Not comparable
ResearchClawBenchSource 14.2%Not comparable
AA Agentic IndexSource 19.9%Not comparable
APEX-Agents-AASource 15.3%Not comparable
GDPval-AASource 23.1%Not comparable
GDPval-AASource 962Not comparable
Coding
BenchmarkMellum2-12B-A2.5B-ThinkingQwen3.5 397BResult
SWE-bench VerifiedSource 76.2%Not comparable
LiveCodeBench v6Source 83.6%Not comparable
SWE-bench ProSource 50.9%Not comparable
AA-SciCodeSource 42.0%Not comparable
AA Coding IndexSource 48.2%Not comparable
Reasoning
BenchmarkMellum2-12B-A2.5B-ThinkingQwen3.5 397BResult
LongBench v2Source 63.2%Not comparable
AI-NeedleSource 68.7%Not comparable
AA-LCRSource 65.7%Not comparable
CritPtSource 1.7%Not comparable
KnowledgeMellum2-12B-A2.5B-Thinking wins
BenchmarkMellum2-12B-A2.5B-ThinkingQwen3.5 397BResult
MMLU-ReduxSource 86.2%94.9%Qwen3.5 397B leads
GPQASource 57.6%88.4%Qwen3.5 397B leads
GPQA-DSource 57.6%Not comparable
SuperGPQASource 70.4%Not comparable
MMLU-ProSource 87.8%Not comparable
C-EvalSource 93%Not comparable
HLESource 28.7%Not comparable
Artificial Analysis Intelligence IndexSource 33.7%Not comparable
AA-GPQA DiamondSource 89.3%Not comparable
AA-HLESource 27.3%Not comparable
AA-Omniscience IndexSource -29.8%Not comparable
AA-Omniscience AccuracySource 31.4%Not comparable
AA-Omniscience Hallucination RateSource 89.1%Not comparable
Math
BenchmarkMellum2-12B-A2.5B-ThinkingQwen3.5 397BResult
AIME26Source 93.3%Not comparable
HMMT Feb 2025Source 94.8%Not comparable
HMMT Nov 2025Source 92.7%Not comparable
HMMT Feb 2026Source 87.9%Not comparable
MMAnswerBenchSource 80.9%Not comparable
Multilingual
BenchmarkMellum2-12B-A2.5B-ThinkingQwen3.5 397BResult
MMLU-ProXSource 84.7%Not comparable
NOVA-63Source 59.1%Not comparable
Multimodal
BenchmarkMellum2-12B-A2.5B-ThinkingQwen3.5 397BResult
MMMU-ProSource 79%Not comparable
MathVisionSource 88.6%Not comparable
CharXivSource 80.8%Not comparable
VideoMMMUSource 84.7%Not comparable
ScreenSpot ProSource 65.6%Not comparable
V*Source 95.8%Not comparable
AA-MMMU-ProSource 77.3%Not comparable
Inst. FollowingQwen3.5 397B wins
BenchmarkMellum2-12B-A2.5B-ThinkingQwen3.5 397BResult
IFEvalSource 76.5%92.6%Qwen3.5 397B leads
AA-IFBenchSource 78.8%Not comparable
Frequently Asked Questions (3)

Which is better, Mellum2-12B-A2.5B-Thinking or Qwen3.5 397B?

Mellum2-12B-A2.5B-Thinking and Qwen3.5 397B are tied on the BenchAlign overall score, so the right pick depends on which category matters most for your use case.

Which is better for knowledge tasks, Mellum2-12B-A2.5B-Thinking or Qwen3.5 397B?

Mellum2-12B-A2.5B-Thinking has the edge for knowledge tasks in this comparison, averaging 57.6 versus 56.6. Inside this category, GPQA is the benchmark that creates the most daylight between them.

Which is better for instruction following, Mellum2-12B-A2.5B-Thinking or Qwen3.5 397B?

Qwen3.5 397B has the edge for instruction following in this comparison, averaging 92.6 versus 76.5. Inside this category, IFEval is the benchmark that creates the most daylight between them.

Related Comparisons

Last updated: July 23, 2026

Choose a model with this week’s evidence

Join 2,000+ readers for ranking moves, pricing changes, and the claims that still need proof.

One email each week. Unsubscribe anytime.