Skip to main content
BenchLM
Advertise

Model comparison

DeepSeek V3 vs Qwen3.5 397B

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

DeepSeek V3

DeepSeek

44.1/100

Supported · Public rank #154

90% interval 25.5–62.8

Qwen3.5 397B

Alibaba

56.2/100

Estimated · Public rank #78

90% interval 44.7–67.7

Qwen3.5 397B has the higher public score estimate, 56.19 versus 44.15, but the 90% score intervals overlap. Treat that as a lead, not a settled winner.

4 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.

  • Chat turn cost

    1K fresh input + 500 output tokens

    DeepSeek V3

    DeepSeek V3 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

    DeepSeek V3

    DeepSeek V3 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

    The category averages use different weighted benchmark sets, so they are directional rather than like-for-like.

    Confidence: limited

  • Agentic work

    Tool use, computer use, and multi-step task completion

    Not enough matched evidence

    No shared weighted benchmark basis supports a winner.

    Confidence: limited

  • Long documents

    Prompts that approach the documented context limit

    No clear pick

    The documented context windows are equal.

    Confidence: documented

  • 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. DeepSeek V3 does not fit this 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.

Shared results
4
DeepSeek V3 only
2
Qwen3.5 397B only
34
Like-for-like categories
1 / 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.

Instruction following

Like-for-like
DeepSeek V3
86.1
Qwen3.5 397B
92.6
Weighted basis
1 vs 1 rows
Reading
Qwen3.5 397B leads

Coding

Directional only
DeepSeek V3
38.9
Qwen3.5 397B
66.5
Weighted basis
2 vs 2 rows
Reading
Directional only

Knowledge

Directional only
DeepSeek V3
72.7
Qwen3.5 397B
56.6
Weighted basis
2 vs 4 rows
Reading
Directional only

Agentic

Not comparable
DeepSeek V3
Not measured
Qwen3.5 397B
56.5
Weighted basis
0 vs 2 rows
Reading
Not comparable

Reasoning

Not comparable
DeepSeek V3
Not measured
Qwen3.5 397B
63.2
Weighted basis
0 vs 1 rows
Reading
Not comparable

Math

Not comparable
DeepSeek V3
1.7
Qwen3.5 397B
90.6
Weighted basis
1 vs 2 rows
Reading
Not comparable

Multilingual

Not comparable
DeepSeek V3
Not measured
Qwen3.5 397B
84.7
Weighted basis
0 vs 1 rows
Reading
Not comparable

Multimodal

Not comparable
DeepSeek V3
Not measured
Qwen3.5 397B
79.6
Weighted basis
0 vs 2 rows
Reading
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.

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

DeepSeek V3
$0.00082
Fits in one request
Qwen3.5 397B
$0.0024
Fits in one request

DeepSeek V3 has the lower modeled cost

Costs use the listed standard API rates.

Repository review

50K fresh input + 3K output tokens

DeepSeek V3
$0.0168
Fits in one request
Qwen3.5 397B
$0.0408
Fits in one request

DeepSeek V3 has the lower modeled cost

Costs use the listed standard API rates.

Cache-heavy agent loop

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

DeepSeek V3
$0.0304
Does not fit in one request
Qwen3.5 397B
$0.168
Does not fit in one request
Cached input priced at the published list-input rate

DeepSeek V3 does not fit this 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.

Specification differences

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

Context window

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

DeepSeek V3

128K

Qwen3.5 397B

128K

API model ID

DeepSeek V3

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.

DeepSeek V3

$0.07 per 1M cached input tokens

Qwen3.5 397B

Not published

Documented inputs

DeepSeek V3

Not sourced

Qwen3.5 397B

Not sourced

Documented outputs

DeepSeek V3

Not sourced

Qwen3.5 397B

Not sourced

Provider availability

DeepSeek V3

Not sourced

Qwen3.5 397B

Not sourced

Reasoning profile

DeepSeek V3

Non-Reasoning

Qwen3.5 397B

Non-Reasoning

Weight access

DeepSeek V3

Open Weight

Qwen3.5 397B

Open Weight

License

DeepSeek V3

Open Weight

Qwen3.5 397B

Open Weight

Release date

DeepSeek V3

2024-12-26

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, 56.19 versus 44.15, but the 90% score intervals overlap.
Workload cost
Repository review: $0.0168 vs $0.0408. Cache-heavy agent loop: $0.0304 vs $0.168.
Context tradeoff
Both models list 128K.

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

Self-host vs API cost

Estimates at 50,000 req/day · 1000 tokens/req average.

DeepSeek V3
API / mo$1,028
Self-host / mo$18,221
Break-even1.2B/day
Qwen3.5 397B
API / mo$3,150
Self-host / moNot listed
Break-even—
Proprietary model — self-hosting not applicable.
Model the full break-even

Benchmark evidence

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

Browse raw public benchmark evidence40 rows

Agentic

  • Terminal-Bench 2.0

    DeepSeek V3—
    Qwen3.5 397B52.5%
    Source

    Not directly comparable

  • BrowseComp

    DeepSeek V3—
    Qwen3.5 397B62%
    Source

    Not directly comparable

  • Claw-Eval

    DeepSeek V3—
    Qwen3.5 397B56.8%
    Source

    Not directly comparable

  • QwenClawBench

    DeepSeek V3—
    Qwen3.5 397B51.8%
    Source

    Not directly comparable

  • τ³-bench results

    DeepSeek V3—
    Qwen3.5 397B68.4%
    Source

    Not directly comparable

  • VITA-Bench

    DeepSeek V3—
    Qwen3.5 397B43.7%
    Source

    Not directly comparable

  • DeepPlanning

    DeepSeek V3—
    Qwen3.5 397B37.6%
    Source

    Not directly comparable

  • Toolathlon

    DeepSeek V3—
    Qwen3.5 397B36.3%
    Source

    Not directly comparable

  • MCP Atlas

    DeepSeek V3—
    Qwen3.5 397B46.1%
    Source

    Not directly comparable

  • MCP-Tasks

    DeepSeek V3—
    Qwen3.5 397B74.2%
    Source

    Not directly comparable

  • WideResearch

    DeepSeek V3—
    Qwen3.5 397B74.0%
    Source

    Not directly comparable

  • Gert Labs

    DeepSeek V3—
    Qwen3.5 397B46.76%
    Source

    Not directly comparable

  • ResearchClawBench

    DeepSeek V3—
    Qwen3.5 397B14.2%
    Source

    Not directly comparable

Coding

  • LiveCodeBench

    DeepSeek V337.6%
    Source
    Qwen3.5 397B—

    Not directly comparable

  • SWE-bench Verified

    DeepSeek V342%
    Source
    Qwen3.5 397B76.2%
    Source

    Qwen3.5 397B leads this result

  • LiveCodeBench v6

    DeepSeek V3—
    Qwen3.5 397B83.6%
    Source

    Not directly comparable

  • SWE-bench Pro

    DeepSeek V3—
    Qwen3.5 397B50.9%
    Source

    Not directly comparable

Reasoning

  • LongBench v2

    DeepSeek V3—
    Qwen3.5 397B63.2%
    Source

    Not directly comparable

  • AI-Needle

    DeepSeek V3—
    Qwen3.5 397B68.7%
    Source

    Not directly comparable

Knowledge

  • GPQA

    DeepSeek V359.1%
    Source
    Qwen3.5 397B88.4%
    Source

    Qwen3.5 397B leads this result

  • MMLU-Pro

    DeepSeek V375.9%
    Source
    Qwen3.5 397B87.8%
    Source

    Qwen3.5 397B leads this result

  • SuperGPQA

    DeepSeek V3—
    Qwen3.5 397B70.4%
    Source

    Not directly comparable

  • MMLU-Redux

    DeepSeek V3—
    Qwen3.5 397B94.9%
    Source

    Not directly comparable

  • C-Eval

    DeepSeek V3—
    Qwen3.5 397B93%
    Source

    Not directly comparable

  • HLE

    DeepSeek V3—
    Qwen3.5 397B28.7%
    Source

    Not directly comparable

Math

  • FrontierMath v2 (Tiers 1-3)

    DeepSeek V31.724%
    Source
    Qwen3.5 397B—

    Not directly comparable

  • AIME26

    DeepSeek V3—
    Qwen3.5 397B93.3%
    Source

    Not directly comparable

  • HMMT Feb 2025

    DeepSeek V3—
    Qwen3.5 397B94.8%
    Source

    Not directly comparable

  • HMMT Nov 2025

    DeepSeek V3—
    Qwen3.5 397B92.7%
    Source

    Not directly comparable

  • HMMT Feb 2026

    DeepSeek V3—
    Qwen3.5 397B87.9%
    Source

    Not directly comparable

  • MMAnswerBench

    DeepSeek V3—
    Qwen3.5 397B80.9%
    Source

    Not directly comparable

Multilingual

  • MMLU-ProX

    DeepSeek V3—
    Qwen3.5 397B84.7%
    Source

    Not directly comparable

  • NOVA-63

    DeepSeek V3—
    Qwen3.5 397B59.1%
    Source

    Not directly comparable

Multimodal

  • MMMU-Pro

    DeepSeek V3—
    Qwen3.5 397B79%
    Source

    Not directly comparable

  • MathVision

    DeepSeek V3—
    Qwen3.5 397B88.6%
    Source

    Not directly comparable

  • CharXiv

    DeepSeek V3—
    Qwen3.5 397B80.8%
    Source

    Not directly comparable

  • VideoMMMU

    DeepSeek V3—
    Qwen3.5 397B84.7%
    Source

    Not directly comparable

  • ScreenSpot Pro

    DeepSeek V3—
    Qwen3.5 397B65.6%
    Source

    Not directly comparable

  • V*

    DeepSeek V3—
    Qwen3.5 397B95.8%
    Source

    Not directly comparable

Instruction following

  • IFEval

    DeepSeek V386.1%
    Source
    Qwen3.5 397B92.6%
    Source

    Qwen3.5 397B leads this result

Frequently asked questions

Which is better, DeepSeek V3 or Qwen3.5 397B?

Qwen3.5 397B has the higher public score estimate, 56.19 versus 44.15, but the 90% score intervals overlap. The higher estimate is not a decisive winner because the uncertainty ranges overlap.

Which is better for coding, DeepSeek V3 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, DeepSeek V3 or Qwen3.5 397B?

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

Which costs less, DeepSeek V3 or Qwen3.5 397B?

For the stated presets, chat costs $0.00082 on DeepSeek V3 and $0.0024 on Qwen3.5 397B; repository review costs $0.0168 and $0.0408; the cache-heavy agent loop costs $0.0304 and $0.168. DeepSeek V3 does not fit this 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.

Which has the larger context window, DeepSeek V3 or Qwen3.5 397B?

Both models list the same context window, 128K.

Related comparisons

Compare API pricing Read the methodology Open the model selector

Last updated July 30, 2026

Watch DeepSeek V3 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.