Model comparison

GPT-5.4 nano vs Qwen3.5 397B

Updated July 31, 2026. Public scores include evidence status and uncertainty. They are not guarantees for a specific workload.

GPT-5.4 nano

OpenAI

66.0/100

Supported · Public rank #28

90% interval 55.5–76.5

Qwen3.5 397B

Alibaba

56.2/100

Estimated · Public rank #79

90% interval 44.7–67.7

GPT-5.4 nano has the higher public score estimate, 65.99 versus 56.22, but the 90% score intervals overlap. Treat that as a lead, not a settled winner.

6 results are shared. Category rows based on different benchmark sets are marked directional and do not name a winner.

Which one for your work

Recommendations appear only when a shared evidence basis or an explicit operating constraint supports the call. Secondary and unsupported use cases stay disclosed below the initial list.

Long documents
Prompts that approach the documented context limit
GPT-5.4 nano
GPT-5.4 nano has the larger documented context window.
Confidence: documented
Chat turn cost
1K fresh input + 500 output tokens
GPT-5.4 nano
GPT-5.4 nano has the lower estimated token cost for this stated workload. Costs use the listed standard API rates.
Confidence: listed-rates

Show secondary and unsupported calls

Repository review cost
50K fresh input + 3K output tokens
GPT-5.4 nano
GPT-5.4 nano has the lower estimated token cost for this stated workload. Costs use the listed standard API rates.
Confidence: listed-rates
Coding work
Code generation, repair, and software-engineering tasks
Not enough matched evidence
No shared weighted benchmark basis supports a winner.
Confidence: limited
Agentic work
Tool use, computer use, and multi-step task completion
Not enough matched evidence
The category averages use different weighted benchmark sets, so they are directional rather than like-for-like.
Confidence: limited
Cache-heavy agent loop cost
200K cached + 20K fresh input + 10K output tokens
Not enough matched evidence
The page does not recommend a cost winner because at least one model cannot fit the stated workload in one request. Qwen3.5 397B does not fit this workload in one request. Qwen3.5 397B has no published cached-input rate, so cached tokens use its listed input rate.
Confidence: rate-fallback

What is actually comparable

Shared results can support a head-to-head reading. Results present for only one model describe coverage, not superiority.

7 GPT-5.4 nano only

6 Shared

32 Qwen3.5 397B only

Shared results: 6
GPT-5.4 nano only: 7
Qwen3.5 397B only: 32
Like-for-like categories: 0 / 8

3 categories use different evidence sets. Those rows remain visible for coverage context but do not name a winner.

Category results, on a stated basis

Each row states whether both averages use the same weighted benchmark set. Directional and not-comparable rows remain visible, but they never receive a winner in this template.

Agentic

Directional only

GPT-5.4 nano: 42.9
Qwen3.5 397B: 56.5
Weighted basis: 2 vs 2 rows
Reading: Directional only

Knowledge

Directional only

GPT-5.4 nano: 43.8
Qwen3.5 397B: 56.6
Weighted basis: 2 vs 4 rows
Reading: Directional only

Multimodal

Directional only

GPT-5.4 nano: 66.1
Qwen3.5 397B: 79.6
Weighted basis: 1 vs 2 rows
Reading: Directional only

Coding

Not comparable

GPT-5.4 nano: Not measured
Qwen3.5 397B: 66.5
Weighted basis: 0 vs 2 rows
Reading: Not comparable

Reasoning

Not comparable

GPT-5.4 nano: Not measured
Qwen3.5 397B: 63.2
Weighted basis: 0 vs 1 rows
Reading: Not comparable

Math

Not comparable

GPT-5.4 nano: 21.0
Qwen3.5 397B: 90.6
Weighted basis: 2 vs 2 rows
Reading: Not comparable

Multilingual

Not comparable

GPT-5.4 nano: Not measured
Qwen3.5 397B: 84.7
Weighted basis: 0 vs 1 rows
Reading: Not comparable

Instruction following

Not comparable

GPT-5.4 nano: Not measured
Qwen3.5 397B: 92.6
Weighted basis: 0 vs 1 rows
Reading: Not comparable

Category averages with the server-provided evidence basis for GPT-5.4 nano and Qwen3.5 397B
Category	GPT-5.4 nano	Qwen3.5 397B	Weighted basis	Reading
Agentic	42.9	56.5	Directional only2 vs 2 rows	Directional only
Knowledge	43.8	56.6	Directional only2 vs 4 rows	Directional only
Multimodal	66.1	79.6	Directional only1 vs 2 rows	Directional only
Coding	Not measured	66.5	Not comparable0 vs 2 rows	Not comparable
Reasoning	Not measured	63.2	Not comparable0 vs 1 rows	Not comparable
Math	21.0	90.6	Not comparable2 vs 2 rows	Not comparable
Multilingual	Not measured	84.7	Not comparable0 vs 1 rows	Not comparable
Instruction following	Not measured	92.6	Not comparable0 vs 1 rows	Not comparable

Shape of the matched evidence

Only shared public evidence is shown. Sparse evidence stays a ruled list rather than being closed into a radar shape.

Too few matched category axes support a radar. The ruled list below shows only shared benchmark results; positions use each benchmark’s normalized display scale when available.

MMMU-Pro
Multimodal
GPT-5.4 nano: 66.1%Qwen3.5 397B: 79%Normalized gap 12.9GPT-5.4 nano source Qwen3.5 397B source
HLE
Knowledge
GPT-5.4 nano: 37.7%Qwen3.5 397B: 28.7%Normalized gap 9.0GPT-5.4 nano source Qwen3.5 397B source
Terminal-Bench 2.0
Agentic
GPT-5.4 nano: 46.3%Qwen3.5 397B: 52.5%Normalized gap 6.2GPT-5.4 nano source Qwen3.5 397B source
GPQA
Knowledge
GPT-5.4 nano: 82.8%Qwen3.5 397B: 88.4%Normalized gap 5.6GPT-5.4 nano source Qwen3.5 397B source

What each workload costs

Three fixed token mixes turn per-token rates into comparable decisions. Each scenario states context fit and whether cached input had to fall back to the published list-input rate.

Chat turn

1K fresh input + 500 output tokens

GPT-5.4 nano: $0.00082; Fits in one request
Qwen3.5 397B: $0.0024; Fits in one request

GPT-5.4 nano has the lower modeled cost

Costs use the listed standard API rates.

Repository review

50K fresh input + 3K output tokens

GPT-5.4 nano: $0.01375; Fits in one request
Qwen3.5 397B: $0.0408; Fits in one request

GPT-5.4 nano has the lower modeled cost

Costs use the listed standard API rates.

Cache-heavy agent loop

200K cached + 20K fresh input + 10K output tokens

GPT-5.4 nano: $0.0205; Fits in one request
Qwen3.5 397B: $0.168; Does not fit in one request; Cached input priced at the published list-input rate

Qwen3.5 397B does not fit this workload in one request. Qwen3.5 397B has no published cached-input rate, so cached tokens use its listed input rate.

Specification differences

Sourced differences are shown directly. Missing facts stay explicit instead of being inferred from a model name or family.

SpecificationGPT-5.4 nanoQwen3.5 397B

Context window

Maximum documented context; output-token limits may be lower.

GPT-5.4 nano

400K

OpenAI GPT-5.4 nano model documentation

Qwen3.5 397B

128K

API model ID

GPT-5.4 nano

gpt-5.4-nano

OpenAI GPT-5.4 nano model documentation

Qwen3.5 397B

Not sourced

Cached-input rate

A missing cached-input rate falls back to the listed input rate only in the stated workload estimate.

GPT-5.4 nano

$0.02 per 1M cached input tokens

OpenAI pricing

Qwen3.5 397B

Not published

Documented inputs

GPT-5.4 nano

text, image

OpenAI model catalog

Qwen3.5 397B

Not sourced

Documented outputs

GPT-5.4 nano

text

OpenAI model catalog

Qwen3.5 397B

Not sourced

Provider availability

GPT-5.4 nano

Generally Available · OpenAI Responses API

OpenAI model catalog

Qwen3.5 397B

Not sourced

Reasoning profile

GPT-5.4 nano

Reasoning

Qwen3.5 397B

Non-Reasoning

Weight access

GPT-5.4 nano

Proprietary

Qwen3.5 397B

Open Weight

License

GPT-5.4 nano

Proprietary

Qwen3.5 397B

Open Weight

Release date

GPT-5.4 nano

2026-03-17

Qwen3.5 397B

2026-02-16

If you already use one of these models

Deployment change: The models list different providers, so authentication, endpoint behavior, limits, and feature support may change.
Quality signal: GPT-5.4 nano has the higher public score estimate, 65.99 versus 56.22, but the 90% score intervals overlap.
Workload cost: Repository review: $0.01375 vs $0.0408. Cache-heavy agent loop: $0.0205 vs $0.168.
Context tradeoff: GPT-5.4 nano has the larger documented window (400K).

Run the same representative tasks against both endpoints before changing production traffic.

Benchmark evidence

The full public result ledger is available for audit without forcing a wide desktop table onto a phone.

Browse raw public benchmark evidence45 rows

Agentic

Terminal-Bench 2.0
GPT-5.4 nano46.3%
Source
Qwen3.5 397B52.5%
Source
Qwen3.5 397B leads this result
OSWorld-Verified
GPT-5.4 nano39%
Source
Qwen3.5 397B—
Not directly comparable
MCP Atlas
GPT-5.4 nano56.1%
Source
Qwen3.5 397B46.1%
Source
GPT-5.4 nano leads this result
Toolathlon
GPT-5.4 nano35.5%
Source
Qwen3.5 397B36.3%
Source
Qwen3.5 397B leads this result
τ²-bench results
GPT-5.4 nano92.5%
Source
Qwen3.5 397B—
Not directly comparable
BrowseComp
GPT-5.4 nano—
Qwen3.5 397B62%
Source
Not directly comparable
Claw-Eval
GPT-5.4 nano—
Qwen3.5 397B56.8%
Source
Not directly comparable
QwenClawBench
GPT-5.4 nano—
Qwen3.5 397B51.8%
Source
Not directly comparable
τ³-bench results
GPT-5.4 nano—
Qwen3.5 397B68.4%
Source
Not directly comparable
VITA-Bench
GPT-5.4 nano—
Qwen3.5 397B43.7%
Source
Not directly comparable
DeepPlanning
GPT-5.4 nano—
Qwen3.5 397B37.6%
Source
Not directly comparable
MCP-Tasks
GPT-5.4 nano—
Qwen3.5 397B74.2%
Source
Not directly comparable
WideResearch
GPT-5.4 nano—
Qwen3.5 397B74.0%
Source
Not directly comparable
Gert Labs
GPT-5.4 nano—
Qwen3.5 397B46.76%
Source
Not directly comparable
ResearchClawBench
GPT-5.4 nano—
Qwen3.5 397B14.2%
Source
Not directly comparable

Coding

Vibe Code Bench
GPT-5.4 nano26.10%
Source
Qwen3.5 397B—
Not directly comparable
SWE-bench Verified
GPT-5.4 nano—
Qwen3.5 397B76.2%
Source
Not directly comparable
LiveCodeBench v6
GPT-5.4 nano—
Qwen3.5 397B83.6%
Source
Not directly comparable
SWE-bench Pro
GPT-5.4 nano—
Qwen3.5 397B50.9%
Source
Not directly comparable

Reasoning

LongBench v2
GPT-5.4 nano—
Qwen3.5 397B63.2%
Source
Not directly comparable
AI-Needle
GPT-5.4 nano—
Qwen3.5 397B68.7%
Source
Not directly comparable

Knowledge

GPQA
GPT-5.4 nano82.8%
Source
Qwen3.5 397B88.4%
Source
Qwen3.5 397B leads this result
HLE
GPT-5.4 nano37.7%
Source
Qwen3.5 397B28.7%
Source
GPT-5.4 nano leads this result
HLE w/o tools
GPT-5.4 nano24.3%
Source
Qwen3.5 397B—
Not directly comparable
SuperGPQA
GPT-5.4 nano—
Qwen3.5 397B70.4%
Source
Not directly comparable
MMLU-Pro
GPT-5.4 nano—
Qwen3.5 397B87.8%
Source
Not directly comparable
MMLU-Redux
GPT-5.4 nano—
Qwen3.5 397B94.9%
Source
Not directly comparable
C-Eval
GPT-5.4 nano—
Qwen3.5 397B93%
Source
Not directly comparable

Math

FrontierMath v2 (Tiers 1-3)
GPT-5.4 nano25.860%
Source
Qwen3.5 397B—
Not directly comparable
FrontierMath v2 (Tier 4)
GPT-5.4 nano6.250%
Source
Qwen3.5 397B—
Not directly comparable
AIME26
GPT-5.4 nano—
Qwen3.5 397B93.3%
Source
Not directly comparable
HMMT Feb 2025
GPT-5.4 nano—
Qwen3.5 397B94.8%
Source
Not directly comparable
HMMT Nov 2025
GPT-5.4 nano—
Qwen3.5 397B92.7%
Source
Not directly comparable
HMMT Feb 2026
GPT-5.4 nano—
Qwen3.5 397B87.9%
Source
Not directly comparable
MMAnswerBench
GPT-5.4 nano—
Qwen3.5 397B80.9%
Source
Not directly comparable

Multilingual

MMLU-ProX
GPT-5.4 nano—
Qwen3.5 397B84.7%
Source
Not directly comparable
NOVA-63
GPT-5.4 nano—
Qwen3.5 397B59.1%
Source
Not directly comparable

Multimodal

MMMU-Pro
GPT-5.4 nano66.1%
Source
Qwen3.5 397B79%
Source
Qwen3.5 397B leads this result
MMMU-Pro w/ Python
GPT-5.4 nano69.5%
Source
Qwen3.5 397B—
Not directly comparable
MathVision
GPT-5.4 nano—
Qwen3.5 397B88.6%
Source
Not directly comparable
CharXiv
GPT-5.4 nano—
Qwen3.5 397B80.8%
Source
Not directly comparable
VideoMMMU
GPT-5.4 nano—
Qwen3.5 397B84.7%
Source
Not directly comparable
ScreenSpot Pro
GPT-5.4 nano—
Qwen3.5 397B65.6%
Source
Not directly comparable
V*
GPT-5.4 nano—
Qwen3.5 397B95.8%
Source
Not directly comparable

Instruction following

IFEval
GPT-5.4 nano—
Qwen3.5 397B92.6%
Source
Not directly comparable

Frequently asked questions

Which is better, GPT-5.4 nano or Qwen3.5 397B?

GPT-5.4 nano has the higher public score estimate, 65.99 versus 56.22, but the 90% score intervals overlap. The higher estimate is not a decisive winner because the uncertainty ranges overlap.

Which is better for coding, GPT-5.4 nano or Qwen3.5 397B?

The published evidence does not provide a shared weighted coding basis for both models, so BenchLM does not name a coding winner.

Which is better for agentic tasks, GPT-5.4 nano or Qwen3.5 397B?

The current agentic tasks averages use different weighted benchmark sets, so BenchLM does not name a winner from them. Read the shared benchmark rows directly and test the models on the same task set.

Which costs less, GPT-5.4 nano or Qwen3.5 397B?

For the stated presets, chat costs $0.00082 on GPT-5.4 nano and $0.0024 on Qwen3.5 397B; repository review costs $0.01375 and $0.0408; the cache-heavy agent loop costs $0.0205 and $0.168. Qwen3.5 397B does not fit this workload in one request. Qwen3.5 397B has no published cached-input rate, so cached tokens use its listed input rate.

Which has the larger context window, GPT-5.4 nano or Qwen3.5 397B?

GPT-5.4 nano has the larger documented context window: 400K, compared with 128K.

Related comparisons

Compare API pricing Read the methodology Open the model selector

Last updated July 31, 2026

Watch GPT-5.4 nano vs Qwen3.5 397B

One weekly email when material rank, price, or benchmark evidence changes make this matchup worth revisiting.

Read a sample issue

Join 2,000+ readers.

GPT-5.4 nano vs Qwen3.5 397B

Which one for your work

Long documents

Chat turn cost

Repository review cost

Coding work

Agentic work

Cache-heavy agent loop cost

What is actually comparable

Category results, on a stated basis

Agentic

Knowledge

Multimodal

Coding

Reasoning

Math

Multilingual

Instruction following

Shape of the matched evidence

What each workload costs

Chat turn

Repository review

Cache-heavy agent loop

Specification differences

Context window

API model ID

Cached-input rate

Documented inputs

Documented outputs

Provider availability

Reasoning profile

Weight access

License

Release date

Benchmark evidence

Agentic

Coding

Reasoning

Knowledge

Math

Multilingual

Multimodal

Instruction following

Frequently asked questions

Which is better, GPT-5.4 nano or Qwen3.5 397B?

Which is better for coding, GPT-5.4 nano or Qwen3.5 397B?

Which is better for agentic tasks, GPT-5.4 nano or Qwen3.5 397B?

Which costs less, GPT-5.4 nano or Qwen3.5 397B?

Which has the larger context window, GPT-5.4 nano or Qwen3.5 397B?

Related comparisons

Watch GPT-5.4 nano vs Qwen3.5 397B