Model A

LiquidAI

—

Evidence status unavailable

90% interval unavailable

LFM2.5-VL-3B vs Qwen3.8-27B

Updated September 10, 2026. Public scores include evidence status and uncertainty. They are not guarantees for a specific workload.

Model B

Qwen3.8-27B

Alibaba

64.52/100

Supported · Public rank #41

90% interval 59.1–69.9

Decision reading

At least one model is not scored in the current public ranking lane, so the page does not name an overall quality winner.

2 results are shared. Category rows resting on Estimated evidence or different benchmark sets are marked directional and do not name a winner.

Share or export

Share on X LinkedInSocial card CSV JSON

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
Qwen3.8-27B
Qwen3.8-27B 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
LFM2.5-VL-3B 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
LFM2.5-VL-3B is not ranked on the public lane for agentic, so no winner is named for agentic.
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. LFM2.5-VL-3B does not fit this workload in one request. LFM2.5-VL-3B has no comparable published API token rate. Qwen3.8-27B has no comparable published API token rate.
Confidence: listed-rates
Repository review cost
50K fresh input + 3K 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. LFM2.5-VL-3B does not fit this workload in one request. LFM2.5-VL-3B has no comparable published API token rate. Qwen3.8-27B has no comparable published API token rate.
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.

8 LFM2.5-VL-3B only

2 Shared

31 Qwen3.8-27B only

Shared results: 2
LFM2.5-VL-3B only: 8
Qwen3.8-27B only: 31
Like-for-like categories: 0 / 8

Category results, on a stated basis

Each row shows the public-lane category score for both models: the BenchAlign 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.

Agentic

Not comparable

LFM2.5-VL-3B: Not ranked
Qwen3.8-27B: 63.4; Supported · #13/152
Basis: BenchAlign lane · 1 vs 8 public rows
Reading: Not comparable

Coding

Not comparable

LFM2.5-VL-3B: Not ranked
Qwen3.8-27B: 53.9; Supported · #43/151
Basis: BenchAlign lane · 0 vs 8 public rows
Reading: Not comparable

Reasoning

Not comparable

LFM2.5-VL-3B: Not ranked
Qwen3.8-27B: 77.4; #7/20
Basis: Provisional lane · 0 vs 0 weighted rows
Reading: Not comparable

Knowledge

Not comparable

LFM2.5-VL-3B: Not ranked
Qwen3.8-27B: 54.1; Supported · #58/183
Basis: BenchAlign lane · 0 vs 6 public rows
Reading: Not comparable

Math

Not comparable

LFM2.5-VL-3B: Not ranked
Qwen3.8-27B: Not ranked
Basis: Provisional lane · 0 vs 0 weighted rows
Reading: Not comparable

Multilingual

Not comparable

LFM2.5-VL-3B: Not ranked
Qwen3.8-27B: Not ranked
Basis: Provisional lane · 0 vs 0 weighted rows
Reading: Not comparable

Multimodal

Not comparable

LFM2.5-VL-3B: Not ranked
Qwen3.8-27B: 80.0; #11/48
Basis: Provisional lane · 1 vs 1 weighted rows
Reading: Not comparable

Instruction following

Not comparable

LFM2.5-VL-3B: Not ranked
Qwen3.8-27B: 84.7; #45/123
Basis: Provisional lane · 1 vs 1 weighted rows
Reading: Not comparable

Category scores with the server-provided evidence basis for LFM2.5-VL-3B and Qwen3.8-27B
Category	LFM2.5-VL-3B	Qwen3.8-27B	Basis	Reading
Agentic	Not ranked	63.4Supported · #13/152	Not comparableBenchAlign lane · 1 vs 8 public rows	Not comparable
Coding	Not ranked	53.9Supported · #43/151	Not comparableBenchAlign lane · 0 vs 8 public rows	Not comparable
Reasoning	Not ranked	77.4#7/20	Not comparableProvisional lane · 0 vs 0 weighted rows	Not comparable
Knowledge	Not ranked	54.1Supported · #58/183	Not comparableBenchAlign lane · 0 vs 6 public rows	Not comparable
Math	Not ranked	Not ranked	Not comparableProvisional lane · 0 vs 0 weighted rows	Not comparable
Multilingual	Not ranked	Not ranked	Not comparableProvisional lane · 0 vs 0 weighted rows	Not comparable
Multimodal	Not ranked	80.0#11/48	Not comparableProvisional lane · 1 vs 1 weighted rows	Not comparable
Instruction following	Not ranked	84.7#45/123	Not comparableProvisional lane · 1 vs 1 weighted rows	Not comparable

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

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.

IFBench
Instruction following
LFM2.5-VL-3B: 25.8%Qwen3.8-27B: 79.5%Normalized gap 53.7LFM2.5-VL-3B source Qwen3.8-27B 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

LFM2.5-VL-3B: Self-hosted; infrastructure cost varies; Fits in one request
Qwen3.8-27B: Self-hosted; infrastructure cost varies; Fits in one request

LFM2.5-VL-3B has no comparable published API token rate. Qwen3.8-27B has no comparable published API token rate.

Repository review

50K fresh input + 3K output tokens

LFM2.5-VL-3B: Self-hosted; infrastructure cost varies; Does not fit in one request
Qwen3.8-27B: Self-hosted; infrastructure cost varies; Fits in one request

LFM2.5-VL-3B does not fit this workload in one request. LFM2.5-VL-3B has no comparable published API token rate. Qwen3.8-27B has no comparable published API token rate.

Cache-heavy agent loop

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

LFM2.5-VL-3B: Self-hosted; infrastructure cost varies; Does not fit in one request; Cached-input rate unavailable
Qwen3.8-27B: Self-hosted; infrastructure cost varies; Fits in one request; Cached-input rate unavailable

LFM2.5-VL-3B does not fit this workload in one request. LFM2.5-VL-3B has no comparable published API token rate. Qwen3.8-27B 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.

SpecificationLFM2.5-VL-3BQwen3.8-27B

Context window

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

LFM2.5-VL-3B

32K

Liquid AI LFM2.5-VL-3B model card

Qwen3.8-27B

262K

Qwen3.8-27B model card

API model ID

LFM2.5-VL-3B

LiquidAI/LFM2.5-VL-3B

Liquid AI LFM2.5-VL-3B model card

Qwen3.8-27B

Not sourced

Cached-input rate

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

LFM2.5-VL-3B

No comparable hosted API rate

Liquid AI LFM2.5-VL-3B model card

Qwen3.8-27B

No comparable hosted API rate

Qwen3.8-27B model card

Documented inputs

LFM2.5-VL-3B

Not sourced

Qwen3.8-27B

Not sourced

Documented outputs

LFM2.5-VL-3B

Not sourced

Qwen3.8-27B

Not sourced

Provider availability

LFM2.5-VL-3B

Not sourced

Qwen3.8-27B

Not sourced

Reasoning profile

LFM2.5-VL-3B

Non-Reasoning

Qwen3.8-27B

Reasoning

Weight access

LFM2.5-VL-3B

Open Weight

Qwen3.8-27B

Open Weight

License

LFM2.5-VL-3B

Open Weight

Qwen3.8-27B

Open Weight

Release date

LFM2.5-VL-3B

2026-08-12

Qwen3.8-27B

2026-08-05

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: At least one model is not scored in the current public ranking lane, so the page does not name an overall quality winner.
Workload cost: A complete comparable API-rate estimate is not available for both models.
Context tradeoff: Qwen3.8-27B has the larger documented window (262K).

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 evidence41 rows

Agentic

BFCL v4
LFM2.5-VL-3B32.5%
Source
Qwen3.8-27B—
Not directly comparable
Terminal-Bench 2.1
LFM2.5-VL-3B—
Qwen3.8-27B73.0%
Source
Not directly comparable
CoWorkBench
LFM2.5-VL-3B—
Qwen3.8-27B70.7%
Source
Not directly comparable
JobBench
LFM2.5-VL-3B—
Qwen3.8-27B33.4%
Source
Not directly comparable
Agents' Last Exam
LFM2.5-VL-3B—
Qwen3.8-27B42.9%
Source
Not directly comparable
OSWorld-Verified
LFM2.5-VL-3B—
Qwen3.8-27B84.3%
Source
Not directly comparable
WebArena-Verified
LFM2.5-VL-3B—
Qwen3.8-27B64.8%
Source
Not directly comparable
AndroidWorld
LFM2.5-VL-3B—
Qwen3.8-27B81.9%
Source
Not directly comparable
Terminal-Bench 2.1 (Vals)
LFM2.5-VL-3B—
Qwen3.8-27B58.4%
Source
Not directly comparable

Coding

Terminal-Bench 2.1
LFM2.5-VL-3B—
Qwen3.8-27B73.0%
Source
Not directly comparable
SWE-bench Pro
LFM2.5-VL-3B—
Qwen3.8-27B61.7%
Source
Not directly comparable
NL2Repo
LFM2.5-VL-3B—
Qwen3.8-27B42.3%
Source
Not directly comparable
DeepSWE
LFM2.5-VL-3B—
Qwen3.8-27B42.2%
Source
Not directly comparable
LiveCodeBench v6
LFM2.5-VL-3B—
Qwen3.8-27B90.3%
Source
Not directly comparable
VulcanBench v3
LFM2.5-VL-3B—
Qwen3.8-27B82.6%
Source
Not directly comparable
LiveCodeBench (Vals)
LFM2.5-VL-3B—
Qwen3.8-27B84.0%
Source
Not directly comparable
SWE-bench (Vals)
LFM2.5-VL-3B—
Qwen3.8-27B86.0%
Source
Not directly comparable

Knowledge

GPQA
LFM2.5-VL-3B—
Qwen3.8-27B89.2%
Source
Not directly comparable
GPQA-D
LFM2.5-VL-3B—
Qwen3.8-27B89.2%
Source
Not directly comparable
HLE
LFM2.5-VL-3B—
Qwen3.8-27B30.8%
Source
Not directly comparable
HLE w/o tools
LFM2.5-VL-3B—
Qwen3.8-27B30.8%
Source
Not directly comparable
GPQA Diamond (Vals)
LFM2.5-VL-3B—
Qwen3.8-27B88.9%
Source
Not directly comparable
MMLU-Pro (Vals)
LFM2.5-VL-3B—
Qwen3.8-27B84.3%
Source
Not directly comparable

Multimodal

RealWorldQA
LFM2.5-VL-3B73.1%
Source
Qwen3.8-27B85.9%
Source
Qwen3.8-27B leads this result
SimpleVQA
LFM2.5-VL-3B35.4%
Source
Qwen3.8-27B—
Not directly comparable
CountBench
LFM2.5-VL-3B87.3%
Source
Qwen3.8-27B—
Not directly comparable
MMMU
LFM2.5-VL-3B48.4%
Source
Qwen3.8-27B—
Not directly comparable
MMMU-Pro
LFM2.5-VL-3B30.5%
Source
Qwen3.8-27B—
Not directly comparable
OCRBench V2
LFM2.5-VL-3B47.5%
Source
Qwen3.8-27B—
Not directly comparable
RefCOCO (avg)
LFM2.5-VL-3B87.9%
Source
Qwen3.8-27B—
Not directly comparable
MathVision
LFM2.5-VL-3B—
Qwen3.8-27B90.0%
Source
Not directly comparable
MathVision w/ Python
LFM2.5-VL-3B—
Qwen3.8-27B94.6%
Source
Not directly comparable
BabyVision
LFM2.5-VL-3B—
Qwen3.8-27B65.7%
Source
Not directly comparable
BabyVision w/ Python
LFM2.5-VL-3B—
Qwen3.8-27B85.6%
Source
Not directly comparable
Vision2Web
LFM2.5-VL-3B—
Qwen3.8-27B62.9%
Source
Not directly comparable
CharXiv w/o tools
LFM2.5-VL-3B—
Qwen3.8-27B83.7%
Source
Not directly comparable
CharXiv
LFM2.5-VL-3B—
Qwen3.8-27B90.2%
Source
Not directly comparable
OmniDocBench 1.5
LFM2.5-VL-3B—
Qwen3.8-27B91.1%
Source
Not directly comparable
ERQA
LFM2.5-VL-3B—
Qwen3.8-27B65.5%
Source
Not directly comparable

Instruction following

IFEval
LFM2.5-VL-3B82.3%
Source
Qwen3.8-27B—
Not directly comparable
IFBench
LFM2.5-VL-3B25.8%
Source
Qwen3.8-27B79.5%
Source
Qwen3.8-27B leads this result

Frequently asked questions

Which is better, LFM2.5-VL-3B or Qwen3.8-27B?

At least one model is not scored in the current public ranking lane, so the page does not name an overall quality winner. The page therefore keeps the decision tied to the specific documented workload.

Which is better for coding, LFM2.5-VL-3B or Qwen3.8-27B?

LFM2.5-VL-3B is not ranked on the public lane for coding, so no winner is named for coding.

Which is better for agentic tasks, LFM2.5-VL-3B or Qwen3.8-27B?

LFM2.5-VL-3B is not ranked on the public lane for agentic tasks, so no winner is named for agentic tasks.

Which costs less, LFM2.5-VL-3B or Qwen3.8-27B?

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, LFM2.5-VL-3B or Qwen3.8-27B?

Qwen3.8-27B has the larger documented context window: 262K, compared with 32K.

Related comparisons

Compare API pricing Read the methodology Open the model selector

Last updated September 10, 2026

Watch LFM2.5-VL-3B vs Qwen3.8-27B

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

Read a sample issue

Join 2,000+ readers.

LFM2.5-VL-3B vs Qwen3.8-27B

Which one for your work

Long documents

Coding work

Agentic work

Chat turn cost

Cache-heavy agent loop cost

Repository review cost

What is actually comparable

Category results, on a stated basis

Agentic

Coding

Reasoning

Knowledge

Math

Multilingual

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

Knowledge

Multimodal

Instruction following

Frequently asked questions

Which is better, LFM2.5-VL-3B or Qwen3.8-27B?

Which is better for coding, LFM2.5-VL-3B or Qwen3.8-27B?

Which is better for agentic tasks, LFM2.5-VL-3B or Qwen3.8-27B?

Which costs less, LFM2.5-VL-3B or Qwen3.8-27B?

Which has the larger context window, LFM2.5-VL-3B or Qwen3.8-27B?

Related comparisons

Watch LFM2.5-VL-3B vs Qwen3.8-27B