Model A

NVIDIA

46.32/100

Estimated · Public rank #162

90% interval 36.5–56.2

Nemotron 3 Ultra vs Qwen3.5 397B

Updated August 30, 2026. Public scores include evidence status and uncertainty. They are not guarantees for a specific workload.

Model B

Qwen3.5 397B

Alibaba

57.78/100

Estimated · Public rank #87

90% interval 46.3–69.3

Decision reading

Qwen3.5 397B has the higher public score estimate, 57.78 versus 46.32, but the 90% score intervals overlap. Treat that as a lead, not a settled winner.

10 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.

Agentic work
Tool use, computer use, and multi-step task completion
Qwen3.5 397B
Qwen3.5 397B leads on the same 2 weighted benchmark rows.
Confidence: limited
Long documents
Prompts that approach the documented context limit
Nemotron 3 Ultra
Nemotron 3 Ultra has the larger documented context window.
Confidence: documented

Show secondary and unsupported calls

Coding work
Code generation, repair, and software-engineering tasks
Not enough matched evidence
The category averages use different weighted benchmark sets, so they are directional rather than like-for-like.
Confidence: limited
Chat turn cost
1K fresh input + 500 output tokens
Not enough matched evidence
A complete comparable API-rate estimate is not available for both models.
Confidence: listed-rates
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. Nemotron 3 Ultra has no comparable published API token rate.
Confidence: rate-fallback
Repository review cost
50K fresh input + 3K output tokens
Not enough matched evidence
A complete comparable API-rate estimate is not available for both models.
Confidence: listed-rates

What is actually comparable

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

9 Nemotron 3 Ultra only

10 Shared

28 Qwen3.5 397B only

Shared results: 10
Nemotron 3 Ultra only: 9
Qwen3.5 397B only: 28
Like-for-like categories: 3 / 8

2 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

Like-for-like

Nemotron 3 Ultra: 51.3
Qwen3.5 397B: 56.5
Weighted basis: 2 vs 2 rows
Reading: Qwen3.5 397B leads

Reasoning

Like-for-like

Nemotron 3 Ultra: 61.9
Qwen3.5 397B: 63.2
Weighted basis: 1 vs 1 rows
Reading: Qwen3.5 397B leads

Multilingual

Like-for-like

Nemotron 3 Ultra: 83.0
Qwen3.5 397B: 84.7
Weighted basis: 1 vs 1 rows
Reading: Qwen3.5 397B leads

Coding

Directional only

Nemotron 3 Ultra: 58.3
Qwen3.5 397B: 66.5
Weighted basis: 2 vs 2 rows
Reading: Directional only

Knowledge

Directional only

Nemotron 3 Ultra: 53.8
Qwen3.5 397B: 56.6
Weighted basis: 3 vs 4 rows
Reading: Directional only

Math

Not comparable

Nemotron 3 Ultra: Not measured
Qwen3.5 397B: 90.6
Weighted basis: 0 vs 2 rows
Reading: Not comparable

Multimodal

Not comparable

Nemotron 3 Ultra: Not measured
Qwen3.5 397B: 79.6
Weighted basis: 0 vs 2 rows
Reading: Not comparable

Instruction following

Not comparable

Nemotron 3 Ultra: 81.7
Qwen3.5 397B: 92.6
Weighted basis: 1 vs 1 rows
Reading: Not comparable

Category averages with the server-provided evidence basis for Nemotron 3 Ultra and Qwen3.5 397B
Category	Nemotron 3 Ultra	Qwen3.5 397B	Weighted basis	Reading
Agentic	51.3	56.5	Like-for-like2 vs 2 rows	Qwen3.5 397B leads
Reasoning	61.9	63.2	Like-for-like1 vs 1 rows	Qwen3.5 397B leads
Multilingual	83.0	84.7	Like-for-like1 vs 1 rows	Qwen3.5 397B leads
Coding	58.3	66.5	Directional only2 vs 2 rows	Directional only
Knowledge	53.8	56.6	Directional only3 vs 4 rows	Directional only
Math	Not measured	90.6	Not comparable0 vs 2 rows	Not comparable
Multimodal	Not measured	79.6	Not comparable0 vs 2 rows	Not comparable
Instruction following	81.7	92.6	Not comparable1 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.

BrowseComp
Agentic
Nemotron 3 Ultra: 44.4%Qwen3.5 397B: 62%Normalized gap 17.6Nemotron 3 Ultra source Qwen3.5 397B source
SWE-bench Verified
Coding
Nemotron 3 Ultra: 71.9%Qwen3.5 397B: 76.2%Normalized gap 4.3Nemotron 3 Ultra source Qwen3.5 397B source
Terminal-Bench 2.0
Agentic
Nemotron 3 Ultra: 56.4%Qwen3.5 397B: 52.5%Normalized gap 3.9Nemotron 3 Ultra source Qwen3.5 397B source
HLE
Knowledge
Nemotron 3 Ultra: 26.7%Qwen3.5 397B: 28.7%Normalized gap 2.0Nemotron 3 Ultra source Qwen3.5 397B source
MMLU-ProX
Multilingual
Nemotron 3 Ultra: 83%Qwen3.5 397B: 84.7%Normalized gap 1.7Nemotron 3 Ultra 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

Nemotron 3 Ultra: Self-hosted; infrastructure cost varies; Fits in one request
Qwen3.5 397B: $0.0024; Fits in one request

Nemotron 3 Ultra has no comparable published API token rate.

Repository review

50K fresh input + 3K output tokens

Nemotron 3 Ultra: Self-hosted; infrastructure cost varies; Fits in one request
Qwen3.5 397B: $0.0408; Fits in one request

Nemotron 3 Ultra has no comparable published API token rate.

Cache-heavy agent loop

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

Nemotron 3 Ultra: Self-hosted; infrastructure cost varies; Fits in one request; Cached-input rate unavailable
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. Nemotron 3 Ultra has no comparable published API token rate.

Specification differences

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

SpecificationNemotron 3 UltraQwen3.5 397B

Context window

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

Nemotron 3 Ultra

Qwen3.5 397B

128K

API model ID

Nemotron 3 Ultra

Not sourced

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.

Nemotron 3 Ultra

No comparable hosted API rate

Qwen3.5 397B

Not published

Documented inputs

Nemotron 3 Ultra

Not sourced

Qwen3.5 397B

Not sourced

Documented outputs

Nemotron 3 Ultra

Not sourced

Qwen3.5 397B

Not sourced

Provider availability

Nemotron 3 Ultra

Not sourced

Qwen3.5 397B

Not sourced

Reasoning profile

Nemotron 3 Ultra

Reasoning

Qwen3.5 397B

Non-Reasoning

Weight access

Nemotron 3 Ultra

Open Weight

Qwen3.5 397B

Open Weight

License

Nemotron 3 Ultra

Open Weight

Qwen3.5 397B

Open Weight

Release date

Nemotron 3 Ultra

2026-06-04

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: Qwen3.5 397B has the higher public score estimate, 57.78 versus 46.32, but the 90% score intervals overlap.
Workload cost: A complete comparable API-rate estimate is not available for both models.
Context tradeoff: Nemotron 3 Ultra has the larger documented window (1M).

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 evidence47 rows

Agentic

Terminal-Bench 2.0
Nemotron 3 Ultra56.4%
Source
Qwen3.5 397B52.5%
Source
Nemotron 3 Ultra leads this result
PinchBench
Nemotron 3 Ultra90.0%
Source
Qwen3.5 397B—
Not directly comparable
BrowseComp
Nemotron 3 Ultra44.4%
Source
Qwen3.5 397B62%
Source
Qwen3.5 397B leads this result
τ³-bench results
Nemotron 3 Ultra70.9%
Source
Qwen3.5 397B68.4%
Source
Nemotron 3 Ultra leads this result
HLE w/ tools
Nemotron 3 Ultra37.4%
Source
Qwen3.5 397B—
Not directly comparable
Claw-Eval
Nemotron 3 Ultra—
Qwen3.5 397B56.8%
Source
Not directly comparable
QwenClawBench
Nemotron 3 Ultra—
Qwen3.5 397B51.8%
Source
Not directly comparable
VITA-Bench
Nemotron 3 Ultra—
Qwen3.5 397B43.7%
Source
Not directly comparable
DeepPlanning
Nemotron 3 Ultra—
Qwen3.5 397B37.6%
Source
Not directly comparable
Toolathlon
Nemotron 3 Ultra—
Qwen3.5 397B36.3%
Source
Not directly comparable
MCP Atlas
Nemotron 3 Ultra—
Qwen3.5 397B46.1%
Source
Not directly comparable
MCP-Tasks
Nemotron 3 Ultra—
Qwen3.5 397B74.2%
Source
Not directly comparable
WideResearch
Nemotron 3 Ultra—
Qwen3.5 397B74.0%
Source
Not directly comparable
Gert Labs
Nemotron 3 Ultra—
Qwen3.5 397B46.76%
Source
Not directly comparable
ResearchClawBench
Nemotron 3 Ultra—
Qwen3.5 397B14.2%
Source
Not directly comparable

Coding

SWE-bench Verified
Nemotron 3 Ultra71.9%
Source
Qwen3.5 397B76.2%
Source
Qwen3.5 397B leads this result
SWE Multilingual
Nemotron 3 Ultra67.7%
Source
Qwen3.5 397B—
Not directly comparable
LiveCodeBench v6
Nemotron 3 Ultra89.0%
Source
Qwen3.5 397B83.6%
Source
Nemotron 3 Ultra leads this result
SciCode
Nemotron 3 Ultra44.6%
Source
Qwen3.5 397B—
Not directly comparable
Terminal-Bench 2.0
Nemotron 3 Ultra56.4%
Source
Qwen3.5 397B—
Not directly comparable
SWE-bench Pro
Nemotron 3 Ultra—
Qwen3.5 397B50.9%
Source
Not directly comparable

Reasoning

CritPt
Nemotron 3 Ultra3.1%
Source
Qwen3.5 397B—
Not directly comparable
LongBench v2
Nemotron 3 Ultra61.9%
Source
Qwen3.5 397B63.2%
Source
Qwen3.5 397B leads this result
AI-Needle
Nemotron 3 Ultra—
Qwen3.5 397B68.7%
Source
Not directly comparable

Knowledge

GPQA
Nemotron 3 Ultra87%
Source
Qwen3.5 397B88.4%
Source
Qwen3.5 397B leads this result
GPQA-D
Nemotron 3 Ultra87.0%
Source
Qwen3.5 397B—
Not directly comparable
HLE
Nemotron 3 Ultra26.7%
Source
Qwen3.5 397B28.7%
Source
Qwen3.5 397B leads this result
HLE w/o tools
Nemotron 3 Ultra26.7%
Source
Qwen3.5 397B—
Not directly comparable
MMLU-Pro
Nemotron 3 Ultra86.8%
Source
Qwen3.5 397B87.8%
Source
Qwen3.5 397B leads this result
SuperGPQA
Nemotron 3 Ultra—
Qwen3.5 397B70.4%
Source
Not directly comparable
MMLU-Redux
Nemotron 3 Ultra—
Qwen3.5 397B94.9%
Source
Not directly comparable
C-Eval
Nemotron 3 Ultra—
Qwen3.5 397B93%
Source
Not directly comparable

Math

AIME26
Nemotron 3 Ultra—
Qwen3.5 397B93.3%
Source
Not directly comparable
HMMT Feb 2025
Nemotron 3 Ultra—
Qwen3.5 397B94.8%
Source
Not directly comparable
HMMT Nov 2025
Nemotron 3 Ultra—
Qwen3.5 397B92.7%
Source
Not directly comparable
HMMT Feb 2026
Nemotron 3 Ultra—
Qwen3.5 397B87.9%
Source
Not directly comparable
MMAnswerBench
Nemotron 3 Ultra—
Qwen3.5 397B80.9%
Source
Not directly comparable

Multilingual

MMLU-ProX
Nemotron 3 Ultra83%
Source
Qwen3.5 397B84.7%
Source
Qwen3.5 397B leads this result
NOVA-63
Nemotron 3 Ultra—
Qwen3.5 397B59.1%
Source
Not directly comparable

Multimodal

MMMU-Pro
Nemotron 3 Ultra—
Qwen3.5 397B79%
Source
Not directly comparable
MathVision
Nemotron 3 Ultra—
Qwen3.5 397B88.6%
Source
Not directly comparable
CharXiv
Nemotron 3 Ultra—
Qwen3.5 397B80.8%
Source
Not directly comparable
VideoMMMU
Nemotron 3 Ultra—
Qwen3.5 397B84.7%
Source
Not directly comparable
ScreenSpot Pro
Nemotron 3 Ultra—
Qwen3.5 397B65.6%
Source
Not directly comparable
V*
Nemotron 3 Ultra—
Qwen3.5 397B95.8%
Source
Not directly comparable

Instruction following

IFBench
Nemotron 3 Ultra81.7%
Source
Qwen3.5 397B—
Not directly comparable
IFEval
Nemotron 3 Ultra—
Qwen3.5 397B92.6%
Source
Not directly comparable

Frequently asked questions

Which is better, Nemotron 3 Ultra or Qwen3.5 397B?

Qwen3.5 397B has the higher public score estimate, 57.78 versus 46.32, but the 90% score intervals overlap. The higher estimate is not a decisive winner because the uncertainty ranges overlap.

Which is better for coding, Nemotron 3 Ultra or Qwen3.5 397B?

The current coding 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 is better for agentic tasks, Nemotron 3 Ultra or Qwen3.5 397B?

Qwen3.5 397B leads the like-for-like agentic tasks comparison across 2 shared weighted benchmark rows.

Which costs less, Nemotron 3 Ultra or Qwen3.5 397B?

Both models do not have comparable published API token rates, so this page does not name a universal price winner.

Which has the larger context window, Nemotron 3 Ultra or Qwen3.5 397B?

Nemotron 3 Ultra has the larger documented context window: 1M, compared with 128K.

Related comparisons

Compare API pricing Read the methodology Open the model selector

Last updated August 30, 2026

Watch Nemotron 3 Ultra 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.

Nemotron 3 Ultra vs Qwen3.5 397B

Which one for your work

Agentic work

Long documents

Coding work

Chat turn cost

Cache-heavy agent loop cost

Repository review cost

What is actually comparable

Category results, on a stated basis

Agentic

Reasoning

Multilingual

Coding

Knowledge

Math

Multimodal

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, Nemotron 3 Ultra or Qwen3.5 397B?

Which is better for coding, Nemotron 3 Ultra or Qwen3.5 397B?

Which is better for agentic tasks, Nemotron 3 Ultra or Qwen3.5 397B?

Which costs less, Nemotron 3 Ultra or Qwen3.5 397B?

Which has the larger context window, Nemotron 3 Ultra or Qwen3.5 397B?

Related comparisons

Watch Nemotron 3 Ultra vs Qwen3.5 397B