Skip to main content

Model comparison

Inkling vs Mellum2-12B-A2.5B-Thinking

Data verified

Head-to-head evidence from 2 shared benchmark results across 1 category. Overall scores shown here use the public BenchAlign v5 ranking lane.

Thinking Machines Lab
69/100
Margin
18.0pts
← winning
1 category wins1 category wins

Verified leaderboard positions: Inkling #17; Mellum2-12B-A2.5B-Thinking unranked

Evidence parity. Inkling and Mellum2-12B-A2.5B-Thinking share 2 comparable benchmark results. 2 of 8 categories are comparable. 14 results are unique to Inkling; 4 to Mellum2-12B-A2.5B-Thinking.

Updated July 15, 2026
Shared results
2
Inkling only
14
Mellum2-12B-A2.5B-Thinking only
4
Comparable categories
2 / 8

Pick Inkling if you want the stronger benchmark profile. Mellum2-12B-A2.5B-Thinking only becomes the better choice if knowledge is the priority or you want the stronger reasoning-first profile.

Confidence note. This is a partial-evidence comparison with 2 shared benchmark results across 1 evidence category; 2 of 8 categories currently have scoreable aggregates for both models. Treat the verdict as directional until coverage is more balanced.

Why this result

Inkling is clearly ahead on the provisional aggregate, 69 to 51. The gap is large enough that you do not need to squint at the spreadsheet to see the difference.

Inkling's sharpest advantage is in instruction following, where it averages 79.8 against 76.5. The single biggest benchmark swing on the page is GPQA, 87.9% to 57.6%. Mellum2-12B-A2.5B-Thinking does hit back in knowledge, so the answer changes if that is the part of the workload you care about most.

Mellum2-12B-A2.5B-Thinking is the reasoning model in the pair, while Inkling 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. Inkling gives you the larger context window at 1M, compared with 128K for Mellum2-12B-A2.5B-Thinking.

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 Inkling and Mellum2-12B-A2.5B-Thinking
CategoryInklingΔMellum2-12B-A2.5B-Thinking
KnowledgeInkling51.7Margin 5.9Mellum2-12B-A2.5B-Thinking57.6
Inst. FollowingInkling79.8Margin 3.3Mellum2-12B-A2.5B-Thinking76.5
AgenticInkling69.4MarginNo overlapMellum2-12B-A2.5B-ThinkingNot measured
CodingInkling68.6MarginNo overlapMellum2-12B-A2.5B-ThinkingNot measured
MathInkling97.1MarginNo overlapMellum2-12B-A2.5B-ThinkingNot measured
MultimodalInkling76.5MarginNo overlapMellum2-12B-A2.5B-ThinkingNot measured

Decisive benchmark drivers

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

More
A · InklingB · Mellum2-12B-A2.5B-Thinking
  1. GPQA

    Knowledge
    Source ↗
    A 87.9%B 57.6%
    Winner: InklingΔ 30.3
    GPQA: Inkling scored 87.9%; Mellum2-12B-A2.5B-Thinking scored 57.6%. Inkling wins this benchmark.

Operational comparison

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

MetricInklingMellum2-12B-A2.5B-ThinkingComparison
Input / output priceUSD per 1M tokensInkling$1.87 input / $4.68 outputMellum2-12B-A2.5B-ThinkingNot availableA complete price comparison is not available.
Generation speedtokens per secondInklingNot availableMellum2-12B-A2.5B-ThinkingNot availableA complete speed comparison is not available.
First-answer latencyseconds to first tokenInklingNot availableMellum2-12B-A2.5B-ThinkingNot availableA complete latency comparison is not available.
Context windowmaximum listed tokensInkling1MMellum2-12B-A2.5B-Thinking128KInkling lists the larger context window.

Benchmark Deep Dive

Agentic
BenchmarkInklingMellum2-12B-A2.5B-ThinkingResult
Terminal-Bench 2.0Source 63.8%Not comparable
BrowseCompSource 77.1%Not comparable
MCP AtlasSource 74.1%Not comparable
Design Arena Agentic Web DevSource 1258Not comparable
BFCL v4Source 45.6%Not comparable
Coding
BenchmarkInklingMellum2-12B-A2.5B-ThinkingResult
SWE-bench VerifiedSource 77.6%Not comparable
SWE-bench ProSource 54.3%Not comparable
Terminal-Bench 2.0Source 63.8%Not comparable
LiveCodeBench v6Source 69.9%Not comparable
KnowledgeMellum2-12B-A2.5B-Thinking wins
BenchmarkInklingMellum2-12B-A2.5B-ThinkingResult
GPQASource 87.9%57.6%Inkling leads
GPQA-DSource 87.9%57.6%Inkling leads
HLESource 46%Not comparable
HLE w/o toolsSource 30%Not comparable
MMLU-ReduxSource 86.2%Not comparable
Math
BenchmarkInklingMellum2-12B-A2.5B-ThinkingResult
AIME26Source 97.1%Not comparable
Multimodal
BenchmarkInklingMellum2-12B-A2.5B-ThinkingResult
MMMU-ProSource 73.5%Not comparable
CharXivSource 82%Not comparable
CharXiv w/o toolsSource 78.1%Not comparable
Inst. FollowingInkling wins
BenchmarkInklingMellum2-12B-A2.5B-ThinkingResult
IFBenchSource 79.8%Not comparable
IFEvalSource 76.5%Not comparable
Frequently Asked Questions (3)

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

Inkling is ahead on BenchLM's provisional leaderboard, 69 to 51. The biggest single separator in this matchup is GPQA, where the scores are 87.9% and 57.6%.

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

Mellum2-12B-A2.5B-Thinking has the edge for knowledge tasks in this comparison, averaging 57.6 versus 51.7. Inside this category, GPQA is the benchmark that creates the most daylight between them.

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

Inkling has the edge for instruction following in this comparison, averaging 79.8 versus 76.5. Mellum2-12B-A2.5B-Thinking stays close enough that the answer can still flip depending on your workload.

Related Comparisons

Last updated: July 15, 2026

The AI models change fast. We track them for you.

A weekly brief for engineers and researchers covering new models, ranking shifts, and pricing changes.

Free. No spam. Unsubscribe anytime.