Skip to main content
BenchLM
Data

LFM3-1.7B vs Qwen3.5 397B

Updated October 27, 2026. Rank says Qwen3.5 397B is ahead. Price, access, and your workload can each overturn that. Public scores include evidence status and uncertainty.

Share or export
Share on XLinkedInAPI/MCP

Decision reading

Qwen3.5 397B has the higher public point estimate, 55.69 versus 22.99. Their conditional score ranges do not overlap. These ranges do not establish rank confidence. 10 results are shared. Category rows resting on Estimated evidence or different benchmark sets are marked directional and do not name a winner.

Model A
LiquidAI logo
LFM3-1.7B

LiquidAI

22.99/100

Estimated · Public rank #211

Conditional range 8.6–37.3

Model B
Alibaba logo
Qwen3.5 397B

Alibaba

55.69/100

Estimated · Public rank #70

Conditional range 41.3–70.0

Shared results
10
LFM3-1.7B only
14
Qwen3.5 397B only
52
Like-for-like categories
1 / 8
Estimated: LFM3-1.7B and Qwen3.5 397B. Conditional ranges do not establish rank confidence.How the comparison works

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

    LFM3-1.7B

    LFM3-1.7B 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

    Qwen3.5 397B is not ranked on the public lane for coding, so no winner is named for coding.

    Confidence: limited

  • Agentic work

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

    Not enough matched evidence

    LFM3-1.7B and Qwen3.5 397B are scored on Estimated evidence for agentic, so the reading is 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. LFM3-1.7B 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

Which one for a specific job

Choose a job from the LLM Selector's task catalog to see the category row it rests on, under the same basis rules as the table below. A directional row stays directional; choosing a job never creates a winner.

The same task catalog as the LLM Selector. Each job names the evidence surface it rests on; nothing here adds a new score.

13.0LFM3-1.7B—Qwen3.5 397B

Not comparable · BenchAlign v5.8

The coding row is not comparable on the public lane: at least one model is not measured or not ranked there.

Coding scores combine specific tasks and setups. Match the editor, harness, and effort to your workflow.

Open the coding leaderboardShortlist models for this job
Same basis rules as the category table below

What is actually comparable

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

2 categories rest on Estimated evidence or different evidence sets. Those rows remain visible for coverage context but do not name a winner.

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.

Bars run 0–100 on each benchmark’s normalized display scale

Category results, on a stated basis

Each row shows the public-lane category score for both models: the BenchAlign v5.8 lane for agentic, coding, and knowledge, and the weighted public rows elsewhere, with the same rank each model page reports. A row is like-for-like only when both scores rest on Supported evidence or the same weighted set. Directional and not-comparable rows remain visible, but they never receive a winner.

Instruction following

Like-for-like
LFM3-1.7B
71.7
#76/156
Qwen3.5 397B
79.9
#63/156
Basis
Provisional lane · 1 vs 1 weighted rows
Reading
Qwen3.5 397B leads

Agentic

Directional only
LFM3-1.7B
14.4
Estimated · #123/134
Qwen3.5 397B
44.9
Estimated · #63/134
Basis
BenchAlign v5.8 lane · 10 vs 14 public rows
Reading
Directional only

Knowledge

Directional only
LFM3-1.7B
25.5
Estimated · #175/189
Qwen3.5 397B
53.2
Estimated · #70/189
Basis
BenchAlign v5.8 lane · 3 vs 6 public rows
Reading
Directional only

Coding

Not comparable
LFM3-1.7B
13.0
Estimated · #142/155
Qwen3.5 397B
Not ranked
Basis
BenchAlign v5.8 lane · 6 vs 6 public rows
Reading
Not comparable

Reasoning

Not comparable
LFM3-1.7B
61.3
Unranked · 1 rankable row
Qwen3.5 397B
74.1
Unranked · 2 rankable rows
Basis
Provisional lane · 0 vs 1 weighted rows
Reading
Not comparable

Multimodal

Not comparable
LFM3-1.7B
Not ranked
Qwen3.5 397B
73.4
#31/58
Basis
Provisional lane · 0 vs 2 weighted rows
Reading
Not comparable

Multilingual

Not comparable
LFM3-1.7B
Not ranked
Qwen3.5 397B
93.8
#5/17
Basis
Provisional lane · 0 vs 1 weighted rows
Reading
Not comparable

Math

Not comparable
LFM3-1.7B
Not ranked
Qwen3.5 397B
81.7
Unranked · 5 rankable rows
Basis
Provisional lane · 0 vs 2 weighted rows
Reading
Not comparable

Ranks count the models scored in each category’s lane, so the agentic, coding, and knowledge denominators (BenchAlign v5.8) differ from the provisional-lane categories. Unranked scores sit on the provisional lane’s scale but never name a winner. Like-for-like rows within 0.5 points read as a practical tie.

Bars run 0–100Methodology

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

LFM3-1.7B
Self-hosted; infrastructure cost varies
Fits in one request
Qwen3.5 397B
$0.0024
Fits in one request

LFM3-1.7B has no comparable published API token rate.

Repository review

50K fresh input + 3K output tokens

LFM3-1.7B
Self-hosted; infrastructure cost varies
Fits in one request
Qwen3.5 397B
$0.0408
Fits in one request

LFM3-1.7B has no comparable published API token rate.

Cache-heavy agent loop

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

LFM3-1.7B
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. LFM3-1.7B has no comparable published API token rate.

Cached input falls back to the list input rate only where a cached rate is unpublished

Specification differences

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

Cached-input rate

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

LFM3-1.7B

No comparable hosted API rate

Liquid AI LFM3-1.7B model card

Qwen3.5 397B

Not published

Documented inputs

LFM3-1.7B

Not sourced

Qwen3.5 397B

Not sourced

Documented outputs

LFM3-1.7B

Not sourced

Qwen3.5 397B

Not sourced

Provider availability

LFM3-1.7B

Not sourced

Qwen3.5 397B

Not sourced

Reasoning profile

LFM3-1.7B

Reasoning

Qwen3.5 397B

Non-Reasoning

Weight access

LFM3-1.7B

Open Weight

Qwen3.5 397B

Open Weight

License

LFM3-1.7B

Open Weight

Qwen3.5 397B

Open Weight

Release date

LFM3-1.7B

2026-10-27

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 point estimate, 55.69 versus 22.99. Their conditional score ranges do not overlap. These ranges do not establish rank confidence.
Workload cost
A complete comparable API-rate estimate is not available for both models.
Context tradeoff
LFM3-1.7B has the larger documented window (256K).
Run the same representative tasks against both endpoints before changing production traffic.

Questions

Which is better, LFM3-1.7B or Qwen3.5 397B?

Qwen3.5 397B has the higher public point estimate, 55.69 versus 22.99. Their conditional score ranges do not overlap. These ranges do not establish rank confidence. The page therefore keeps the decision tied to the specific documented workload.

Which is better for coding, LFM3-1.7B or Qwen3.5 397B?

Qwen3.5 397B is not ranked on the public lane for coding, so no winner is named for coding.

Which is better for agentic tasks, LFM3-1.7B or Qwen3.5 397B?

Qwen3.5 397B scores higher for agentic tasks on the public lane, 44.9 to 14.4. LFM3-1.7B and Qwen3.5 397B are scored on Estimated evidence for agentic tasks, so the reading is directional rather than like-for-like. BenchLM does not name a winner for agentic tasks from a directional reading; read the shared benchmark rows directly and test the models on the same task set.

Which costs less, LFM3-1.7B 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, LFM3-1.7B or Qwen3.5 397B?

LFM3-1.7B has the larger documented context window: 256K, compared with 128K.

Self-host vs API cost

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

LFM3-1.7B
API / mo$0
Self-host / mo$298
Break-even—
Qwen3.5 397B
API / mo$3,150
Self-host / mo$13,981
Break-even311M/day
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 evidence76 rows

Agentic

  • AutomationBench

    LFM3-1.7B41.3%
    Source
    Qwen3.5 397B—

    Not directly comparable

  • BFCL v4

    LFM3-1.7B71.4%
    Source
    Qwen3.5 397B—

    Not directly comparable

  • τ³-bench results

    LFM3-1.7B68.0%
    Source
    Qwen3.5 397B68.4%
    Source

    Qwen3.5 397B leads this result

  • WebArena-Verified

    LFM3-1.7B38.4%
    Source
    Qwen3.5 397B—

    Not directly comparable

  • MCP Atlas

    LFM3-1.7B32.6%
    Source
    Qwen3.5 397B46.1%
    Source

    Qwen3.5 397B leads this result

  • ToolSandbox

    LFM3-1.7B80.36%
    Source
    Qwen3.5 397B—

    Not directly comparable

  • Claw-Eval

    LFM3-1.7B68.6%
    Source
    Qwen3.5 397B56.8%
    Source

    LFM3-1.7B leads this result

  • HLE w/ tools

    LFM3-1.7B13.9%
    Source
    Qwen3.5 397B—

    Not directly comparable

  • Terminal-Bench 2.1

    LFM3-1.7B12.9%
    Source
    Qwen3.5 397B—

    Not directly comparable

  • LiveClaw

    LFM3-1.7B45.60%
    Source
    Qwen3.5 397B—

    Not directly comparable

  • Terminal-Bench 2.0

    LFM3-1.7B—
    Qwen3.5 397B52.5%
    Source

    Not directly comparable

  • BrowseComp

    LFM3-1.7B—
    Qwen3.5 397B69%
    Source

    Not directly comparable

  • OSWorld-Verified

    LFM3-1.7B—
    Qwen3.5 397B62.2%
    Source

    Not directly comparable

  • QwenClawBench

    LFM3-1.7B—
    Qwen3.5 397B51.8%
    Source

    Not directly comparable

  • VITA-Bench

    LFM3-1.7B—
    Qwen3.5 397B43.7%
    Source

    Not directly comparable

  • DeepPlanning

    LFM3-1.7B—
    Qwen3.5 397B37.6%
    Source

    Not directly comparable

  • Toolathlon

    LFM3-1.7B—
    Qwen3.5 397B36.3%
    Source

    Not directly comparable

  • MCP-Tasks

    LFM3-1.7B—
    Qwen3.5 397B74.2%
    Source

    Not directly comparable

  • WideResearch

    LFM3-1.7B—
    Qwen3.5 397B74.0%
    Source

    Not directly comparable

  • Gert Labs

    LFM3-1.7B—
    Qwen3.5 397B46.76%
    Source

    Not directly comparable

  • ResearchClawBench

    LFM3-1.7B—
    Qwen3.5 397B14.2%
    Source

    Not directly comparable

Coding

  • SciCode

    LFM3-1.7B28.5%
    Source
    Qwen3.5 397B—

    Not directly comparable

  • SWE-bench Verified

    LFM3-1.7B43.4%
    Source
    Qwen3.5 397B76.2%
    Source

    Qwen3.5 397B leads this result

  • SWE Multilingual

    LFM3-1.7B23%
    Source
    Qwen3.5 397B69.3%
    Source

    Qwen3.5 397B leads this result

  • Terminal-Bench 2.1

    LFM3-1.7B12.9%
    Source
    Qwen3.5 397B—

    Not directly comparable

  • LiveCodeBench v6

    LFM3-1.7B73.5%
    Source
    Qwen3.5 397B83.6%
    Source

    Qwen3.5 397B leads this result

  • LiveCodeBench Pro

    LFM3-1.7B31.7%
    Source
    Qwen3.5 397B—

    Not directly comparable

  • SWE-bench Pro

    LFM3-1.7B—
    Qwen3.5 397B50.9%
    Source

    Not directly comparable

  • NL2Repo

    LFM3-1.7B—
    Qwen3.5 397B32.2%
    Source

    Not directly comparable

  • Terminal-Bench 2.0

    LFM3-1.7B—
    Qwen3.5 397B52.5%
    Source

    Not directly comparable

Reasoning

  • RULER (128K)

    LFM3-1.7B65.56%
    Source
    Qwen3.5 397B—

    Not directly comparable

  • LongBench v2

    LFM3-1.7B—
    Qwen3.5 397B63.2%
    Source

    Not directly comparable

  • AI-Needle

    LFM3-1.7B—
    Qwen3.5 397B68.7%
    Source

    Not directly comparable

Multimodal

  • MMMU

    LFM3-1.7B—
    Qwen3.5 397B85.0%
    Source

    Not directly comparable

  • MMMU-Pro

    LFM3-1.7B—
    Qwen3.5 397B79%
    Source

    Not directly comparable

  • RealWorldQA

    LFM3-1.7B—
    Qwen3.5 397B83.9%
    Source

    Not directly comparable

  • OmniDocBench 1.5

    LFM3-1.7B—
    Qwen3.5 397B90.8%
    Source

    Not directly comparable

  • Video-MME (with subtitle)

    LFM3-1.7B—
    Qwen3.5 397B87.5%
    Source

    Not directly comparable

  • Video-MME (w/o subtitle)

    LFM3-1.7B—
    Qwen3.5 397B83.7%
    Source

    Not directly comparable

  • MathVision

    LFM3-1.7B—
    Qwen3.5 397B88.6%
    Source

    Not directly comparable

  • We-Math

    LFM3-1.7B—
    Qwen3.5 397B87.9%
    Source

    Not directly comparable

  • DynaMath

    LFM3-1.7B—
    Qwen3.5 397B86.3%
    Source

    Not directly comparable

  • SimpleVQA

    LFM3-1.7B—
    Qwen3.5 397B67.1%
    Source

    Not directly comparable

  • CharXiv

    LFM3-1.7B—
    Qwen3.5 397B80.8%
    Source

    Not directly comparable

  • MMLongBench-Doc

    LFM3-1.7B—
    Qwen3.5 397B61.5%
    Source

    Not directly comparable

  • CC-OCR

    LFM3-1.7B—
    Qwen3.5 397B82.0%
    Source

    Not directly comparable

  • AI2D_TEST

    LFM3-1.7B—
    Qwen3.5 397B93.9%
    Source

    Not directly comparable

  • CountBench

    LFM3-1.7B—
    Qwen3.5 397B97.2%
    Source

    Not directly comparable

  • RefCOCO (avg)

    LFM3-1.7B—
    Qwen3.5 397B92.3%
    Source

    Not directly comparable

  • ODINW13

    LFM3-1.7B—
    Qwen3.5 397B47.0%
    Source

    Not directly comparable

  • ERQA

    LFM3-1.7B—
    Qwen3.5 397B67.5%
    Source

    Not directly comparable

  • VideoMMMU

    LFM3-1.7B—
    Qwen3.5 397B84.7%
    Source

    Not directly comparable

  • MLVU (M-Avg)

    LFM3-1.7B—
    Qwen3.5 397B86.7%
    Source

    Not directly comparable

  • ScreenSpot Pro

    LFM3-1.7B—
    Qwen3.5 397B65.6%
    Source

    Not directly comparable

  • V*

    LFM3-1.7B—
    Qwen3.5 397B95.8%
    Source

    Not directly comparable

Knowledge

  • GPQA

    LFM3-1.7B63.4%
    Source
    Qwen3.5 397B88.4%
    Source

    Qwen3.5 397B leads this result

  • GPQA-D

    LFM3-1.7B63.4%
    Source
    Qwen3.5 397B—

    Not directly comparable

  • HLE

    LFM3-1.7B7.2%
    Source
    Qwen3.5 397B28.7%
    Source

    Qwen3.5 397B leads this result

  • SuperGPQA

    LFM3-1.7B—
    Qwen3.5 397B70.4%
    Source

    Not directly comparable

  • MMLU-Pro

    LFM3-1.7B—
    Qwen3.5 397B87.8%
    Source

    Not directly comparable

  • MMLU-Redux

    LFM3-1.7B—
    Qwen3.5 397B94.9%
    Source

    Not directly comparable

  • C-Eval

    LFM3-1.7B—
    Qwen3.5 397B93%
    Source

    Not directly comparable

Multilingual

  • MMLU-ProX

    LFM3-1.7B—
    Qwen3.5 397B84.7%
    Source

    Not directly comparable

  • NOVA-63

    LFM3-1.7B—
    Qwen3.5 397B59.1%
    Source

    Not directly comparable

  • INCLUDE

    LFM3-1.7B—
    Qwen3.5 397B85.6%
    Source

    Not directly comparable

  • PolyMath

    LFM3-1.7B—
    Qwen3.5 397B73.3%
    Source

    Not directly comparable

  • MAXIFE

    LFM3-1.7B—
    Qwen3.5 397B88.2%
    Source

    Not directly comparable

Instruction following

  • IFEval

    LFM3-1.7B96.7%
    Source
    Qwen3.5 397B92.6%
    Source

    LFM3-1.7B leads this result

  • IFBench

    LFM3-1.7B73.0%
    Source
    Qwen3.5 397B76.5%
    Source

    Qwen3.5 397B leads this result

  • Multi-IF

    LFM3-1.7B89.10%
    Source
    Qwen3.5 397B—

    Not directly comparable

  • IFStruct

    LFM3-1.7B93.72%
    Source
    Qwen3.5 397B—

    Not directly comparable

Math

  • AIME26

    LFM3-1.7B—
    Qwen3.5 397B93.3%
    Source

    Not directly comparable

  • HMMT Feb 2025

    LFM3-1.7B—
    Qwen3.5 397B94.8%
    Source

    Not directly comparable

  • HMMT Nov 2025

    LFM3-1.7B—
    Qwen3.5 397B92.7%
    Source

    Not directly comparable

  • HMMT Feb 2026

    LFM3-1.7B—
    Qwen3.5 397B87.9%
    Source

    Not directly comparable

  • MMAnswerBench

    LFM3-1.7B—
    Qwen3.5 397B80.9%
    Source

    Not directly comparable

76 public results · 10 shared

Related comparisons

Compare API pricingRead the methodologyOpen the model selector

Watch LFM3-1.7B 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 6,100+ subscribers.

Last updated October 27, 2026