Why is this comparison indexable?

It has 3 verifiable comparison points, and both models have pricing or benchmark data.

Are missing metrics invented?

No. Metrics without LMSpeed data are omitted from this report.

Back to models

Data points: 92

Key Takeaways

The readout for GPT-5.4 and gpt-oss-20b, before the detailed comparison sheet.

Decision read

GPT-5.4

GPT-5.4 currently has the stronger profile, with verified wins split 3 to 0.

Evidence depth

92 data points

Includes 35 benchmark rows, 0 audit samples, and 0 provider examples.

Selection signal

Start with GPT-5.4

The charts below split 35 high-signal samples across speed, scores, and audit health.

Change comparison

Switch either side of this report to compare another model with the same LMSpeed data pipeline.

Model AModel B

Select a different model to open a new comparison URL.

Model compare GPT-5.4 vs gpt-oss-20b gpt-5-4-vs-gpt-oss-20b	Model A GPT-5.4	Model B gpt-oss-20b
Overall leader	Leading	Contender
Verified metric wins	3 wins	0 wins
Where it leads	Free providers, Provider coverage, Recent tests

Model compare

GPT-5.4 vs gpt-oss-20b

gpt-5-4-vs-gpt-oss-20b

Model A

GPT-5.4

Model B

gpt-oss-20b

Overall leader

Leading

Contender

Verified metric wins

3 wins

0 wins

Where it leads

Free providers, Provider coverage, Recent tests

Benchmark score comparison

Third-party benchmark profile synced into LMSpeed; only metrics available for both models are shown.

Category performance

Compare benchmark category scores on a 0-100 scale. Select a category to inspect the gap.

Model A coverage: 6 / 8
Model B coverage: 6 / 8
Shared: 6 shared categories

GPT-5.4gpt-oss-20b

Avg. score

GPT-5.4

63.6

Avg. score

gpt-oss-20b

Selected category

Agents

GPT-5.4 leads by 30.5

Professional benchmark details

Metric-level scores with benchmark source, rank depth, confidence, error, and evaluation date where available.

Group

Aggregateverified

BenchLM overall score

SourceGPT-5.4

GPT-5.4

winner

86.0

Rank #3/96 · confidence 4 · eval date 2026-03-05

+67.0

gpt-oss-20b

19.0

Rank #87/96 · confidence 1 · eval date 2025-08-05

Performanceverified

Time to first token

SourceGPT-5.4 (Non-reasoning)

GPT-5.4

0.69 s

Rank #24/113 · confidence 4

+0.19 s

gpt-oss-20b

winner

0.50 s

Rank #6/113 · confidence 4

Performanceverified

Output speed

SourceGPT-5.4 (Non-reasoning)

GPT-5.4

106.7 tok/s

Rank #47/113 · confidence 4

+118.8 tok/s

gpt-oss-20b

winner

225.5 tok/s

Rank #8/113 · confidence 4

Performanceverified

Time to first answer token

SourceGPT-5.4 (Non-reasoning)

GPT-5.4

winner

0.69 s

Rank #19/113 · confidence 4

+8.68 s

gpt-oss-20b

9.37 s

Rank #63/113 · confidence 4

Pricingverified

Blended price

SourceGPT-5.4 (Non-reasoning)

GPT-5.4

$5.63/M

Rank #116/138 · confidence 4

+$5.53/M

gpt-oss-20b

winner

$0.095/M

Rank #3/138 · confidence 4

Pricingverified

Input price

SourceGPT-5.4 (Non-reasoning)

GPT-5.4

$2.50/M

Rank #114/138 · confidence 4

+$2.44/M

gpt-oss-20b

winner

$0.060/M

Rank #4/138 · confidence 4

Pricingverified

Output price

SourceGPT-5.4 (Non-reasoning)

GPT-5.4

$15.00/M

Rank #116/138 · confidence 4

+$14.80/M

gpt-oss-20b

winner

$0.200/M

Rank #4/138 · confidence 4

Reasoningverified

CritPt

SourceGPT-5.4

GPT-5.4

winner

23.4

Rank #2/54 · confidence 4 · eval date 2026-03-05

+22.0

gpt-oss-20b

1.4

Rank #33/54 · confidence 1 · eval date 2025-08-05

Reasoningverified

AA-LCR

SourceGPT-5.4

GPT-5.4

winner

74.0

Rank #4/67 · confidence 4 · eval date 2026-03-05

+43.3

gpt-oss-20b

30.7

Rank #60/67 · confidence 1 · eval date 2025-08-05

Reasoningverified

BenchLM Reasoning score

SourceGPT-5.4

GPT-5.4

winner

87.5

Rank #4/65 · confidence 4 · eval date 2026-03-05

+75.1

gpt-oss-20b

12.4

Rank #64/65 · confidence 1 · eval date 2025-08-05

Reasoningverified

HLE

SourceGPT-5.4 (Non-reasoning)

GPT-5.4

winner

10.6%

Rank #68/150 · confidence 4

+5.5%

gpt-oss-20b

5.1%

Rank #105/150 · confidence 4

Reasoningverified

GPQA

SourceGPT-5.4 (Non-reasoning)

GPT-5.4

winner

74.8%

Rank #75/152 · confidence 4

+13.7%

gpt-oss-20b

61.1%

Rank #110/152 · confidence 4

Codingverified

React Native Evals

SourceGPT-5.4

GPT-5.4

winner

85.3

Rank #1/11 · confidence 4 · eval date 2026-03-05

+14.3

gpt-oss-20b

71.0

Rank #11/11 · confidence 1 · eval date 2025-08-05

Codingverified

AA-SciCode

SourceGPT-5.4

GPT-5.4

winner

56.6

Rank #2/74 · confidence 4 · eval date 2026-03-05

+22.2

gpt-oss-20b

34.4

Rank #60/74 · confidence 1 · eval date 2025-08-05

Codingverified

Terminal-Bench Hard

SourceGPT-5.4

GPT-5.4

winner

57.6

Rank #2/69 · confidence 4 · eval date 2026-03-05

+47.0

gpt-oss-20b

10.6

Rank #60/69 · confidence 1 · eval date 2025-08-05

Codingverified

AA Coding Index

SourceGPT-5.4

GPT-5.4

winner

71.0

Rank #3/36 · confidence 4 · eval date 2026-03-05

+50.3

gpt-oss-20b

20.7

Rank #30/36 · confidence 1 · eval date 2025-08-05

Codingverified

BenchLM Coding score

SourceGPT-5.4

GPT-5.4

winner

86.4

Rank #5/83 · confidence 4 · eval date 2026-03-05

+71.9

gpt-oss-20b

14.5

Rank #81/83 · confidence 1 · eval date 2025-08-05

Codingverified

SciCode

SourceGPT-5.4 (Non-reasoning)

GPT-5.4

winner

47.1%

Rank #13/149 · confidence 4

+13.1%

gpt-oss-20b

34.0%

Rank #92/149 · confidence 4

Knowledgeverified

BenchLM Knowledge score

SourceGPT-5.4

GPT-5.4

winner

96.9

Rank #1/79 · confidence 4 · eval date 2026-03-05

+86.4

gpt-oss-20b

10.5

Rank #79/79 · confidence 1 · eval date 2025-08-05

Knowledgeverified

AA-GPQA Diamond

SourceGPT-5.4

GPT-5.4

winner

92.0

Rank #2/74 · confidence 4 · eval date 2026-03-05

+23.2

gpt-oss-20b

68.8

Rank #56/74 · confidence 1 · eval date 2025-08-05

Knowledgeverified

AA-Omniscience Hallucination Rate

SourceGPT-5.4

GPT-5.4

88.6

Rank #15/68 · confidence 4 · eval date 2026-03-05

+5.5

gpt-oss-20b

winner

94.1

Rank #2/68 · confidence 1 · eval date 2025-08-05

Knowledgeverified

AA-HLE

SourceGPT-5.4

GPT-5.4

winner

41.6

Rank #3/74 · confidence 4 · eval date 2026-03-05

+31.8

gpt-oss-20b

9.8

Rank #51/74 · confidence 1 · eval date 2025-08-05

Knowledgeverified

Artificial Analysis Intelligence Index

SourceGPT-5.4

GPT-5.4

winner

51.4

Rank #3/77 · confidence 4 · eval date 2026-03-05

+36.5

gpt-oss-20b

14.9

Rank #63/77 · confidence 1 · eval date 2025-08-05

Knowledgeverified

AA-Omniscience Accuracy

SourceGPT-5.4

GPT-5.4

winner

50.0

Rank #4/68 · confidence 4 · eval date 2026-03-05

+34.5

gpt-oss-20b

15.5

Rank #63/68 · confidence 1 · eval date 2025-08-05

Multilingualverified

BenchLM Multilingual score

SourceGPT-5.4

GPT-5.4

winner

100.0

Rank #1/59 · confidence 4 · eval date 2026-03-05

+94.4

gpt-oss-20b

5.6

Rank #53/59 · confidence 1 · eval date 2025-08-05

Multimodalverified

Design Arena Website

SourceGPT-5.4

GPT-5.4

winner

1264.0

Rank #21/50 · confidence 4 · eval date 2026-03-05

+369.0

gpt-oss-20b

895.0

Rank #48/50 · confidence 1 · eval date 2025-08-05

Multimodalverified

BenchLM Multimodal Grounded score

SourceGPT-5.4

GPT-5.4

winner

60.0

Rank #41/70 · confidence 4 · eval date 2026-03-05

+51.2

gpt-oss-20b

8.8

Rank #70/70 · confidence 1 · eval date 2025-08-05

Instruction followingverified

BenchLM Instruction Following score

SourceGPT-5.4

GPT-5.4

winner

92.6

Rank #1/66 · confidence 4 · eval date 2026-03-05

+72.3

gpt-oss-20b

20.3

Rank #66/66 · confidence 1 · eval date 2025-08-05

Instruction followingverified

AA-IFBench

SourceGPT-5.4

GPT-5.4

winner

73.9

Rank #18/70 · confidence 4 · eval date 2026-03-05

+8.8

gpt-oss-20b

65.1

Rank #36/70 · confidence 1 · eval date 2025-08-05

Agentsverified

APEX-Agents-AA

SourceGPT-5.4

GPT-5.4

winner

33.3

Rank #2/16 · confidence 4 · eval date 2026-03-05

+32.6

gpt-oss-20b

0.7

Rank #16/16 · confidence 1 · eval date 2025-08-05

Agentsverified

BenchLM Agentic score

SourceGPT-5.4

GPT-5.4

winner

88.3

Rank #3/81 · confidence 4 · eval date 2026-03-05

+70.7

gpt-oss-20b

17.6

Rank #75/81 · confidence 1 · eval date 2025-08-05

Agentsverified

AA Agentic Index

SourceGPT-5.4

GPT-5.4

winner

41.1

Rank #4/32 · confidence 4 · eval date 2026-03-05

+38.0

gpt-oss-20b

3.1

Rank #27/32 · confidence 1 · eval date 2025-08-05

Agentsverified

GDPval-AA

SourceGPT-5.4

GPT-5.4

winner

44.8

Rank #4/27 · confidence 4 · eval date 2026-03-05

+42.1

gpt-oss-20b

2.7

Rank #27/27 · confidence 1 · eval date 2025-08-05

Agentsverified

GDPval-AA

SourceGPT-5.4

GPT-5.4

winner

1395.0

Rank #5/33 · confidence 4 · eval date 2026-03-05

+841.0

gpt-oss-20b

554.0

Rank #28/33 · confidence 1 · eval date 2025-08-05

Agentsverified

Tau2-Telecom

SourceGPT-5.4

GPT-5.4

winner

87.1

Rank #27/68 · confidence 4 · eval date 2026-03-05

+26.9

gpt-oss-20b

60.2

Rank #50/68 · confidence 1 · eval date 2025-08-05

API audit comparison

Latest completed audits from shared providers, with four safety and integrity score groups plus report links.

Provider	GPT-5.4	gpt-oss-20b
No completed audits are available from shared providers yet.

Provider examples

Speed aggregates and input/output pricing share each provider row for real API selection and migration cost checks.

Provider	GPT-5.4	gpt-oss-20b
No shared provider examples are available yet.

When to choose each model

This report only uses LMSpeed data for GPT-5.4 and gpt-oss-20b: pricing, speed aggregates, third-party benchmark scores, and shared provider samples.

Guidance	GPT-5.4	gpt-oss-20b
When to choose each model	GPT-5.4 GPT-5.4 is stronger when you prioritize Free providers, Provider coverage, Recent tests.	gpt-oss-20b gpt-oss-20b does not clearly lead on the verified metrics here, so check provider-specific pricing before choosing.

FAQ

TL;DR: GPT-5.4 leads across 92 verifiable data points, including pricing, speed, latency, benchmarks, and provider examples.

Why is this comparison indexable?: It has 3 verifiable comparison points, and both models have pricing or benchmark data.
Are missing metrics invented?: No. Metrics without LMSpeed data are omitted from this report.

Continue from GPT-5.4 vs gpt-oss-20b into nearby model comparisons with enough verified LMSpeed data.

Model metadata	GPT-5.4 exposes 1.1M tokens; notable signals: Text input, Image input, File input, Text output.	gpt-oss-20b exposes 131.1K tokens; notable signals: Text input, Text output, Tool calling, Structured outputs.
Developer	OpenAI	OpenAI
Context window	1.1M tokens	131.1K tokens
Max output	128K tokens	No data
Released	Mar 2026	No data
Modalities	Input TextImageFile Output Text	Input Text Output Text
Features	Text inputImage inputFile inputText outputTool callingStructured outputsJSON modeReasoning	Text inputText outputTool callingStructured outputsJSON modeReasoning
Parameters	No data	20B
Tokenizer	GPT	GPT
Knowledge cutoff	No data	2024-06-30
OpenRouter ID	openai/gpt-5.4	openai/gpt-oss-20b
References	No data	No data

Key Takeaways

Change comparison

GPT-5.4 vs gpt-oss-20b

GPT-5.4

gpt-oss-20b

Benchmark score comparison

Category performance

Professional benchmark details

BenchLM overall score

Time to first token

Output speed

Time to first answer token

Blended price

Input price

Output price

CritPt

AA-LCR

BenchLM Reasoning score

HLE

GPQA

React Native Evals

AA-SciCode

Terminal-Bench Hard

AA Coding Index

BenchLM Coding score

SciCode

BenchLM Knowledge score

AA-GPQA Diamond

AA-Omniscience Hallucination Rate

AA-HLE

Artificial Analysis Intelligence Index

AA-Omniscience Accuracy

BenchLM Multilingual score

Design Arena Website

BenchLM Multimodal Grounded score

BenchLM Instruction Following score

AA-IFBench

APEX-Agents-AA

BenchLM Agentic score

AA Agentic Index

GDPval-AA

GDPval-AA

Tau2-Telecom

API audit comparison

Provider examples

When to choose each model

FAQ

Related compare reports