Noometry

Model comparison

GPT-5.4 vs Qwen3 235B-A22B

GPT-5.4 is the stronger model overall, scoring 59.4 to 43.5 on the Noometry Index. Qwen3 235B-A22B costs 4.6× less per token, which makes it the better buy when GPT-5.4's lead doesn't matter for your workload.

Last verified . 37 shared benchmarks.

GPT-5.4 OpenAI

59.4

Rank #16 Confirmed

Qwen3 235B-A22B Alibaba (Qwen)

43.5

Rank #91 Confirmed

Summary

  • They share 37 benchmarks with published results for both. GPT-5.4 scores higher in 9 categories and Qwen3 235B-A22B in 0 categories; 9 gaps are clear of the uncertainty.
  • The widest gap is in reasoning, where GPT-5.4 leads 61.8 to 15.7.
  • The biggest single-benchmark swing is ARC-AGI-1: 93.7% for GPT-5.4 and 11% for Qwen3 235B-A22B.
  • Qwen3 235B-A22B is cheaper at $0.70 / $2.80 per million input/output tokens, against $2.50 / $15 for GPT-5.4.
  • GPT-5.4 accepts more context: 1.05M tokens versus 131K.
  • Qwen3 235B-A22B has downloadable open weights; the other is API-only.

Side by side

GPT-5.4Qwen3 235B-A22B
ProviderOpenAIAlibaba (Qwen)
Noometry Index59.443.5
Released2026-03-052025-04
WeightsProprietaryOpen
Context window1.05M131K
Max output128K16K
Input $ / M tokens$2.50$0.70
Output $ / M tokens$15$2.80
Results tracked6849

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

Category by category

Coding GPT-5.4 leads

GPT-5.4: 52.6 (#33), Qwen3 235B-A22B: 44.3 (#75)

BenchmarkGPT-5.4Qwen3 235B-A22B
SciCode56.6%42.4%
WeirdML77.7%41%
LMArena Coding14971445
SWE-bench Verified76.9%—
DeepSWE51.8%—
Aider Polyglot—59.6%
LMArena WebDev1465—
GSO31.4%—
MirrorCode15.6%—
ALE-Bench1,607—
AlgoTune1.85—

Agentic & Tool Use GPT-5.4 leads

GPT-5.4: 46.5 (#13), Qwen3 235B-A22B: 33.9 (#51)

BenchmarkGPT-5.4Qwen3 235B-A22B
Vending-Bench 26,144-11.34
Terminal-Bench81.8%—
APEX-Agents52.4%—
Berkeley Function Calling Leaderboard—52.1%
τ²-bench Banking39.4%—
DeepResearch Bench35.1%—
PostTrainBench19%—
GBAEval45.1%—
LMArena Search1197—
METR Time Horizons74.3%—

Reasoning GPT-5.4 leads

GPT-5.4: 61.8 (#19), Qwen3 235B-A22B: 15.7 (#311)

BenchmarkGPT-5.4Qwen3 235B-A22B
ARC-AGI-274%1.3%
Kagi LLM Benchmark63.8%69.4%
ARC-AGI-193.7%11%
CritPt23.4%0%
Chess Puzzles44%12%
LMArena Hard Prompts14851433
Mystery Game Puzzles37%9%
DTBench94.4%80.3%
LMCA52%29.3%
Epoch Capabilities Index156.81143.85
ForecastBench59.559.7
SimpleBench—31%
NYT Connections (extended)91.3%—
EnigmaEval16%—
Thematic Generalization80%—
EBR-Bench25.4%—

Math GPT-5.4 leads

GPT-5.4: 73.5 (#19), Qwen3 235B-A22B: 50.4 (#57)

BenchmarkGPT-5.4Qwen3 235B-A22B
OTIS Mock AIME 2024-202597.8%86.7%
LMArena Math14881432
FrontierMath (Feb 2025 set)47.6%8.5%
FrontierMath Tier 4 (v1)27.1%0%
FrontierMath (Tiers 1-3)78.6%—
FrontierMath Tier 449%—
MathArena Final-Answer Competitions83.1%—
ProofBench56%—
Omni-MATH—71.8%
MATH Level 5—68.9%

Knowledge GPT-5.4 leads

GPT-5.4: 65.3 (#14), Qwen3 235B-A22B: 49.6 (#73)

BenchmarkGPT-5.4Qwen3 235B-A22B
GPQA Diamond93.3%80.1%
SimpleQA Verified45.1%40.4%
Vectara Hallucination Rate7%9.3%
LMArena Expert15071463
Humanity's Last Exam36.2%—
MMLU-Pro—84.4%
Confabulations—15.6%
GPQA (HELM)—72.7%

Multimodal Not comparable

GPT-5.4: 43.7 (#20), Qwen3 235B-A22B: —

BenchmarkGPT-5.4Qwen3 235B-A22B
LMArena Vision1303—
Blueprint-Bench 227.1%—
Furniture Assembly37.5%—
LMArena Document1471—

Multilingual GPT-5.4 leads

GPT-5.4: 56.2 (#23), Qwen3 235B-A22B: 52.3 (#89)

BenchmarkGPT-5.4Qwen3 235B-A22B
LMArena Non-English14651409
LMArena Chinese15191481
LMArena French14931445
LMArena German14721433
LMArena Japanese14851399
LMArena Korean14481391
LMArena Russian14801411
LMArena Spanish14541430

Instruction Following GPT-5.4 leads

GPT-5.4: 77.1 (#27), Qwen3 235B-A22B: 72.6 (#136)

BenchmarkGPT-5.4Qwen3 235B-A22B
LMArena Instruction Following14691408
IFEval—83.5%

Long Context GPT-5.4 leads

GPT-5.4: 50.3 (#8), Qwen3 235B-A22B: 46.1 (#26)

BenchmarkGPT-5.4Qwen3 235B-A22B
LMArena Longer Query14731426
Fiction.LiveBench—75%
CL-bench27.9%—
CL-bench Life21.7%—

Writing & Preference GPT-5.4 leads

GPT-5.4: 71.9 (#17), Qwen3 235B-A22B: 59.6 (#108)

BenchmarkGPT-5.4Qwen3 235B-A22B
LMArena Text14691419
LMArena Creative Writing14391384
EQ-Bench Creative Writing18401366
LMArena Multi-Turn14821432
Short-Story Creative Writing—83%
WildBench—86.6%
EQ-Bench 41272—

Frequently asked questions

Is GPT-5.4 better than Qwen3 235B-A22B?

GPT-5.4 is the stronger model overall, scoring 59.4 to 43.5 on the Noometry Index. Qwen3 235B-A22B costs 4.6× less per token, which makes it the better buy when GPT-5.4's lead doesn't matter for your workload.

Which is cheaper, GPT-5.4 or Qwen3 235B-A22B?

Qwen3 235B-A22B is cheaper. It lists at $0.70 per million input tokens and $2.80 per million output tokens; GPT-5.4 lists at $2.50 and $15.

Is GPT-5.4 or Qwen3 235B-A22B better for coding?

GPT-5.4 scores higher on coding benchmarks: 52.6 versus 44.3 in the Noometry coding category.

Which has the bigger context window?

GPT-5.4 does, with 1.05M tokens against 131K.

How many benchmarks do GPT-5.4 and Qwen3 235B-A22B share?

37 benchmarks have published results for both models. GPT-5.4 has 68 scored results on Noometry and Qwen3 235B-A22B has 49.

Related comparisons

Go deeper