Noometry

Model comparison

GPT-4 vs Qwen3 14B

Qwen3 14B is the stronger model overall, scoring 35.5 to 29.1 on the Noometry Index.

Last verified . 6 shared benchmarks.

GPT-4 OpenAI

29.1

Rank #316 Confirmed

Qwen3 14B Alibaba (Qwen)

35.5

Rank #225 Confirmed

Summary

  • They share 6 benchmarks with published results for both. GPT-4 scores higher in 0 categories and Qwen3 14B in 5 categories; 3 gaps are clear of the uncertainty.
  • The widest gap is in math, where Qwen3 14B leads 38.6 to 10.8.
  • The biggest single-benchmark swing is OTIS Mock AIME 2024-2025: 1.1% for GPT-4 and 66.4% for Qwen3 14B.
  • Qwen3 14B is cheaper at $0.35 / $1.40 per million input/output tokens, against $30 / $60 for GPT-4.
  • Qwen3 14B accepts more context: 131K tokens versus 8K.
  • Qwen3 14B has downloadable open weights; the other is API-only.

Side by side

GPT-4Qwen3 14B
ProviderOpenAIAlibaba (Qwen)
Noometry Index29.135.5
Released2023-03-142025-04
WeightsProprietaryOpen
Context window8K131K
Max output8K8K
Input $ / M tokens$30$0.35
Output $ / M tokens$60$1.40
Results tracked3812

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

Category by category

Coding Qwen3 14B leads

GPT-4: 31.6 (#283), Qwen3 14B: 37.3 (#195)

BenchmarkGPT-4Qwen3 14B
SciCode—31.6%
WeirdML12.4%—
BigCodeBench Instruct46%—
LMArena Coding1254—
BigCodeBench Complete57.2%—
HumanEval+79.3%—

Agentic & Tool Use Not comparable

GPT-4: —, Qwen3 14B: 29.6 (#83)

BenchmarkGPT-4Qwen3 14B
Berkeley Function Calling Leaderboard—41%
METR Time Horizons36.1%—

Reasoning Too close to call

GPT-4: 17.8 (#289), Qwen3 14B: 18.5 (#280)

BenchmarkGPT-4Qwen3 14B
Chess Puzzles4%4%
DTBench62.7%64%
LMCA17.1%18.2%
Epoch Capabilities Index125.89138.23
Kagi LLM Benchmark—49.1%
CritPt—0%
LMArena Hard Prompts1241—
Mystery Game Puzzles12%—
BIG-Bench Hard75.1%—
ForecastBench57.8—
HellaSwag95.3%—
WinoGrande87.5%—

Math Qwen3 14B leads

GPT-4: 10.8 (#309), Qwen3 14B: 38.6 (#133)

BenchmarkGPT-4Qwen3 14B
OTIS Mock AIME 2024-20251.1%66.4%
LMArena Math1269—
MATH Level 523%—
GSM8K92%—

Knowledge Qwen3 14B leads

GPT-4: 18.4 (#282), Qwen3 14B: 39.3 (#134)

BenchmarkGPT-4Qwen3 14B
GPQA Diamond35.7%63.8%
Vectara Hallucination Rate—5.4%
LMArena Expert1211—
MMLU86.4%—
TriviaQA84.8%—

Multilingual Not comparable

GPT-4: 40.6 (#215), Qwen3 14B: —

BenchmarkGPT-4Qwen3 14B
LMArena Non-English1246—
LMArena Chinese1242—
LMArena French1283—
LMArena German1251—
LMArena Japanese1209—
LMArena Korean1184—
LMArena Russian1251—
LMArena Spanish1261—

Instruction Following Not comparable

GPT-4: 65.3 (#222), Qwen3 14B: —

BenchmarkGPT-4Qwen3 14B
LMArena Instruction Following1241—

Long Context Too close to call

GPT-4: 37.7 (#212), Qwen3 14B: 38.1 (#204)

BenchmarkGPT-4Qwen3 14B
Fiction.LiveBench—62.5%
LMArena Longer Query1244—

Writing & Preference Not comparable

GPT-4: 34.9 (#268), Qwen3 14B: —

BenchmarkGPT-4Qwen3 14B
LMArena Text1263—
LMArena Creative Writing1244—
EQ-Bench Creative Writing752—
LMArena Multi-Turn1257—

Frequently asked questions

Is GPT-4 better than Qwen3 14B?

Qwen3 14B is the stronger model overall, scoring 35.5 to 29.1 on the Noometry Index.

Which is cheaper, GPT-4 or Qwen3 14B?

Qwen3 14B is cheaper. It lists at $0.35 per million input tokens and $1.40 per million output tokens; GPT-4 lists at $30 and $60.

Is GPT-4 or Qwen3 14B better for coding?

Qwen3 14B scores higher on coding benchmarks: 37.3 versus 31.6 in the Noometry coding category.

Which has the bigger context window?

Qwen3 14B does, with 131K tokens against 8K.

How many benchmarks do GPT-4 and Qwen3 14B share?

6 benchmarks have published results for both models. GPT-4 has 38 scored results on Noometry and Qwen3 14B has 12.

Related comparisons

Go deeper