Noometry

Model comparison

Inkling vs Qwen3 14B

Inkling is the stronger model overall, scoring 44.1 to 35.5 on the Noometry Index. Qwen3 14B costs 4.2× less per token, which makes it the better buy when Inkling's lead doesn't matter for your workload.

Last verified . 8 shared benchmarks.

Inkling Thinking Machines Lab

44.1

Rank #80 Confirmed

Qwen3 14B Alibaba (Qwen)

35.5

Rank #225 Confirmed

Summary

  • They share 8 benchmarks with published results for both. Inkling scores higher in 3 categories and Qwen3 14B in 3 categories; 5 gaps are clear of the uncertainty.
  • The widest gap is in reasoning, where Inkling leads 40.4 to 18.5.
  • The biggest single-benchmark swing is GPQA Diamond: 88.3% for Inkling and 63.8% for Qwen3 14B.
  • Qwen3 14B is cheaper at $0.35 / $1.40 per million input/output tokens, against $1.87 / $4.68 for Inkling.
  • Qwen3 14B accepts more context: 131K tokens versus 66K.

Side by side

InklingQwen3 14B
ProviderThinking Machines LabAlibaba (Qwen)
Noometry Index44.135.5
Released2026-07-152025-04
WeightsOpenOpen
Context window66K131K
Max output66K8K
Input $ / M tokens$1.87$0.35
Output $ / M tokens$4.68$1.40
Results tracked4112

Sponsored placements are available on pages like this one. Advertise on Noometry

Category by category

Coding Qwen3 14B leads

Inkling: 34.5 (#234), Qwen3 14B: 37.3 (#195)

BenchmarkInklingQwen3 14B
SciCode47%31.6%
FrontierCode14%—
LMArena WebDev1413—
FrontierSWE4.1%—
WeirdML32.3%—
LMArena Coding1464—
ALE-Bench946—

Agentic & Tool Use Too close to call

Inkling: 29.6 (#85), Qwen3 14B: 29.6 (#83)

BenchmarkInklingQwen3 14B
APEX-Agents33.8%—
Berkeley Function Calling Leaderboard—41%
τ²-bench Banking25%—

Reasoning Inkling leads

Inkling: 40.4 (#56), Qwen3 14B: 18.5 (#280)

BenchmarkInklingQwen3 14B
CritPt5.4%0%
Chess Puzzles21%4%
DTBench87.5%64%
LMCA37.6%18.2%
Epoch Capabilities Index148.54138.23
ARC-AGI-236.5%—
SimpleBench50%—
Kagi LLM Benchmark—49.1%
ARC-AGI-179.5%—
LMArena Hard Prompts1451—

Math Qwen3 14B leads

Inkling: 31.3 (#225), Qwen3 14B: 38.6 (#133)

BenchmarkInklingQwen3 14B
OTIS Mock AIME 2024-202588.9%66.4%
FrontierMath (Tiers 1-3)33.3%—
FrontierMath Tier 44.9%—
ProofBench0%—
LMArena Math1479—

Knowledge Inkling leads

Inkling: 55.1 (#49), Qwen3 14B: 39.3 (#134)

BenchmarkInklingQwen3 14B
GPQA Diamond88.3%63.8%
SimpleQA Verified40.3%—
Vectara Hallucination Rate—5.4%
LMArena Expert1465—

Multilingual Not comparable

Inkling: 54.0 (#52), Qwen3 14B: —

BenchmarkInklingQwen3 14B
LMArena Non-English1434—
LMArena Chinese1490—
LMArena French1458—
LMArena German1446—
LMArena Japanese1429—
LMArena Korean1404—
LMArena Russian1429—
LMArena Spanish1448—

Instruction Following Not comparable

Inkling: 75.1 (#71), Qwen3 14B: —

BenchmarkInklingQwen3 14B
LMArena Instruction Following1426—

Long Context Inkling leads

Inkling: 43.8 (#86), Qwen3 14B: 38.1 (#204)

BenchmarkInklingQwen3 14B
Fiction.LiveBench—62.5%
LMArena Longer Query1434—

Writing & Preference Not comparable

Inkling: 65.2 (#51), Qwen3 14B: —

BenchmarkInklingQwen3 14B
LMArena Text1441—
LMArena Creative Writing1387—
EQ-Bench Creative Writing1611—
EQ-Bench 41226—
LMArena Multi-Turn1436—

Frequently asked questions

Is Inkling better than Qwen3 14B?

Inkling is the stronger model overall, scoring 44.1 to 35.5 on the Noometry Index. Qwen3 14B costs 4.2× less per token, which makes it the better buy when Inkling's lead doesn't matter for your workload.

Which is cheaper, Inkling or Qwen3 14B?

Qwen3 14B is cheaper. It lists at $0.35 per million input tokens and $1.40 per million output tokens; Inkling lists at $1.87 and $4.68.

Is Inkling or Qwen3 14B better for coding?

Qwen3 14B scores higher on coding benchmarks: 37.3 versus 34.5 in the Noometry coding category.

Which has the bigger context window?

Qwen3 14B does, with 131K tokens against 66K.

How many benchmarks do Inkling and Qwen3 14B share?

8 benchmarks have published results for both models. Inkling has 41 scored results on Noometry and Qwen3 14B has 12.

Related comparisons

Go deeper