Model comparison

GPT-5.4 vs Qwen2.5-Coder-32B

GPT-5.4 is the stronger model overall, scoring 59.4 to 33.4 on the Noometry Index. Qwen2.5-Coder-32B costs 7.6× less per token, which makes it the better buy when GPT-5.4's lead doesn't matter for your workload.

Last verified October 10, 2026. 13 shared benchmarks.

GPT-5.4 OpenAI

59.4

Rank #16 Confirmed

Qwen2.5-Coder-32B

Alibaba (Qwen)

33.4

Rank #245 Confirmed

Summary

They share 13 benchmarks with published results for both. GPT-5.4 scores higher in 8 categories and Qwen2.5-Coder-32B in 0 categories; 8 gaps are clear of the uncertainty.
The widest gap is in reasoning, where GPT-5.4 leads 61.8 to 21.2.
Qwen2.5-Coder-32B is cheaper at $0.66 / $1 per million input/output tokens, against $2.50 / $15 for GPT-5.4.
GPT-5.4 accepts more context: 1.05M tokens versus 33K.
Qwen2.5-Coder-32B has downloadable open weights; the other is API-only.

Side by side

GPT-5.4 and Qwen2.5-Coder-32B specifications
	GPT-5.4	Qwen2.5-Coder-32B
Provider	OpenAI	Alibaba (Qwen)
Noometry Index	59.4	33.4
Released	2026-03-05	2024-09-18
Weights	Proprietary	Open
Context window	1.05M	33K
Max output	128K	29K
Input $ / M tokens	$2.50	$0.66
Output $ / M tokens	$15	$1
Results tracked	68	31

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), Qwen2.5-Coder-32B: 22.6 (#333)

Coding benchmarks
Benchmark	GPT-5.4	Qwen2.5-Coder-32B
LMArena Coding	1497	1276
SWE-bench Verified	76.9%	—
DeepSWE	51.8%	—
SWE-bench Verified (bash only)	—	9%
Aider Polyglot	—	16.4%
LMArena WebDev	1465	—
SciCode	56.6%	—
GSO	31.4%	—
WeirdML	77.7%	—
BigCodeBench Instruct	—	49%
LiveBench Coding	—	56.9%
MirrorCode	15.6%	—
BigCodeBench Complete	—	58%
ALE-Bench	1,607	—
AlgoTune	1.85	—
HumanEval+	—	87.2%
MBPP+	—	77%

Agentic & Tool Use Not comparable

GPT-5.4: 46.5 (#13), Qwen2.5-Coder-32B: —

Agentic & Tool Use benchmarks
Benchmark	GPT-5.4	Qwen2.5-Coder-32B
Terminal-Bench	81.8%	—
APEX-Agents	52.4%	—
τ²-bench Banking	39.4%	—
DeepResearch Bench	35.1%	—
PostTrainBench	19%	—
GBAEval	45.1%	—
LMArena Search	1197	—
METR Time Horizons	74.3%	—
Vending-Bench 2	6,144	—

Reasoning GPT-5.4 leads

GPT-5.4: 61.8 (#19), Qwen2.5-Coder-32B: 21.2 (#225)

Reasoning benchmarks
Benchmark	GPT-5.4	Qwen2.5-Coder-32B
LMArena Hard Prompts	1485	1251
Epoch Capabilities Index	156.81	119.49
ARC-AGI-2	74%	—
Kagi LLM Benchmark	63.8%	—
NYT Connections (extended)	91.3%	—
ARC-AGI-1	93.7%	—
CritPt	23.4%	—
Chess Puzzles	44%	—
EnigmaEval	16%	—
Thematic Generalization	80%	—
EBR-Bench	25.4%	—
LiveBench Reasoning	—	42.1%
Mystery Game Puzzles	37%	—
DTBench	94.4%	—
LiveBench Data Analysis	—	49.9%
LMCA	52%	—
ForecastBench	59.5	—
HellaSwag	—	83%
LiveBench	—	46.2%
WinoGrande	—	80.8%

Math GPT-5.4 leads

GPT-5.4: 73.5 (#19), Qwen2.5-Coder-32B: 33.3 (#204)

Math benchmarks
Benchmark	GPT-5.4	Qwen2.5-Coder-32B
LMArena Math	1488	1251
FrontierMath (Tiers 1-3)	78.6%	—
FrontierMath Tier 4	49%	—
MathArena Final-Answer Competitions	83.1%	—
OTIS Mock AIME 2024-2025	97.8%	—
ProofBench	56%	—
LiveBench Math	—	46.6%
FrontierMath (Feb 2025 set)	47.6%	—
FrontierMath Tier 4 (v1)	27.1%	—
GSM8K	—	93%

Knowledge GPT-5.4 leads

GPT-5.4: 65.3 (#14), Qwen2.5-Coder-32B: 33.4 (#203)

Knowledge benchmarks
Benchmark	GPT-5.4	Qwen2.5-Coder-32B
LMArena Expert	1507	1221
GPQA Diamond	93.3%	—
Humanity's Last Exam	36.2%	—
SimpleQA Verified	45.1%	—
Vectara Hallucination Rate	7%	—
ARC (AI2) Challenge	—	70.5%
MMLU	—	79.1%

Multimodal Not comparable

GPT-5.4: 43.7 (#20), Qwen2.5-Coder-32B: —

Multimodal benchmarks
Benchmark	GPT-5.4	Qwen2.5-Coder-32B
LMArena Vision	1303	—
Blueprint-Bench 2	27.1%	—
Furniture Assembly	37.5%	—
LMArena Document	1471	—

Multilingual GPT-5.4 leads

GPT-5.4: 56.2 (#23), Qwen2.5-Coder-32B: 37.8 (#235)

Multilingual benchmarks
Benchmark	GPT-5.4	Qwen2.5-Coder-32B
LMArena Non-English	1465	1205
LMArena Chinese	1519	1222
LMArena Russian	1480	1228
LMArena French	1493	—
LMArena German	1472	—
LMArena Japanese	1485	—
LMArena Korean	1448	—
LMArena Spanish	1454	—

Instruction Following GPT-5.4 leads

GPT-5.4: 77.1 (#27), Qwen2.5-Coder-32B: 61.4 (#245)

Instruction Following benchmarks
Benchmark	GPT-5.4	Qwen2.5-Coder-32B
LMArena Instruction Following	1469	1223
LiveBench Instruction Following	—	58.7%

Long Context GPT-5.4 leads

GPT-5.4: 50.3 (#8), Qwen2.5-Coder-32B: 38.0 (#208)

Long Context benchmarks
Benchmark	GPT-5.4	Qwen2.5-Coder-32B
LMArena Longer Query	1473	1251
CL-bench	27.9%	—
CL-bench Life	21.7%	—

Writing & Preference GPT-5.4 leads

GPT-5.4: 71.9 (#17), Qwen2.5-Coder-32B: 41.6 (#240)

Writing & Preference benchmarks
Benchmark	GPT-5.4	Qwen2.5-Coder-32B
LMArena Text	1469	1230
LMArena Creative Writing	1439	1174
LMArena Multi-Turn	1482	1222
EQ-Bench Creative Writing	1840	—
EQ-Bench 4	1272	—
LiveBench Language	—	23.3%

Frequently asked questions

Is GPT-5.4 better than Qwen2.5-Coder-32B?

Which is cheaper, GPT-5.4 or Qwen2.5-Coder-32B?

Qwen2.5-Coder-32B is cheaper. It lists at $0.66 per million input tokens and $1 per million output tokens; GPT-5.4 lists at $2.50 and $15.

Is GPT-5.4 or Qwen2.5-Coder-32B better for coding?

GPT-5.4 scores higher on coding benchmarks: 52.6 versus 22.6 in the Noometry coding category.

Which has the bigger context window?

GPT-5.4 does, with 1.05M tokens against 33K.

How many benchmarks do GPT-5.4 and Qwen2.5-Coder-32B share?

13 benchmarks have published results for both models. GPT-5.4 has 68 scored results on Noometry and Qwen2.5-Coder-32B has 31.