LFM2.5-8B-A1B vs Qwen3.8 Max

Decision reading

Qwen3.8 Max has the higher public score, 71.76 versus 39.15, and the 90% score intervals do not overlap.

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

Model A

LFM2.5-8B-A1B

LiquidAI

39.15/100

Estimated · Public rank #193

90% interval 27.6–50.7

Model B

Qwen3.8 Max

Alibaba

71.76/100

Supported · Public rank #11

90% interval 68.1–75.4

Updated September 15, 2026. Rank says Qwen3.8 Max is ahead. Price, access, and your workload can each overturn that. Public scores include evidence status and uncertainty.

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 Max
Qwen3.8 Max 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
LFM2.5-8B-A1B 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
LFM2.5-8B-A1B is 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. LFM2.5-8B-A1B does not fit this workload in one request. LFM2.5-8B-A1B has no comparable published API token rate. Qwen3.8 Max has no comparable published API token rate.
Confidence: listed-rates
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.

6 LFM2.5-8B-A1B only

1 Shared

59 Qwen3.8 Max only

Shared results: 1
LFM2.5-8B-A1B only: 6
Qwen3.8 Max only: 59
Like-for-like categories: 2 / 8

1 category rests on Estimated evidence or different evidence sets. Those rows remain visible for coverage context but do not name a winner.

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.

Knowledge

Like-for-like

LFM2.5-8B-A1B: 34.5; Supported · #163/183
Qwen3.8 Max: 68.8; Supported · #17/183
Basis: BenchAlign lane · 0 vs 6 public rows
Reading: Qwen3.8 Max leads

Instruction following

Like-for-like

LFM2.5-8B-A1B: 44.7; #94/123
Qwen3.8 Max: 90.7; #18/123
Basis: Provisional lane · 1 vs 1 weighted rows
Reading: Qwen3.8 Max leads

Agentic

Directional only

LFM2.5-8B-A1B: 43.5; Estimated · #96/153
Qwen3.8 Max: 67.3; Supported · #9/153
Basis: BenchAlign lane · 2 vs 15 public rows
Reading: Directional only

Coding

Not comparable

LFM2.5-8B-A1B: Not ranked
Qwen3.8 Max: 60.8; Supported · #20/152
Basis: BenchAlign lane · 0 vs 12 public rows
Reading: Not comparable

Reasoning

Not comparable

LFM2.5-8B-A1B: 21.0; Unranked · 2 rankable rows
Qwen3.8 Max: 86.5; Unranked · 2 rankable rows
Basis: Provisional lane · 0 vs 2 weighted rows
Reading: Not comparable

Math

Not comparable

LFM2.5-8B-A1B: 27.7; Unranked · 3 rankable rows
Qwen3.8 Max: Not ranked
Basis: Provisional lane · 1 vs 0 weighted rows
Reading: Not comparable

Multilingual

Not comparable

LFM2.5-8B-A1B: Not ranked
Qwen3.8 Max: Not ranked
Basis: Provisional lane · 0 vs 0 weighted rows
Reading: Not comparable

Multimodal

Not comparable

LFM2.5-8B-A1B: Not ranked
Qwen3.8 Max: 87.4; #5/48
Basis: Provisional lane · 0 vs 2 weighted rows
Reading: Not comparable

Category scores with the server-provided evidence basis for LFM2.5-8B-A1B and Qwen3.8 Max
Category	LFM2.5-8B-A1B	Qwen3.8 Max	Basis	Reading
Knowledge	34.5Supported · #163/183	68.8Supported · #17/183	Like-for-likeBenchAlign lane · 0 vs 6 public rows	Qwen3.8 Max leads
Instruction following	44.7#94/123	90.7#18/123	Like-for-likeProvisional lane · 1 vs 1 weighted rows	Qwen3.8 Max leads
Agentic	43.5Estimated · #96/153	67.3Supported · #9/153	Directional onlyBenchAlign lane · 2 vs 15 public rows	Directional only
Coding	Not ranked	60.8Supported · #20/152	Not comparableBenchAlign lane · 0 vs 12 public rows	Not comparable
Reasoning	21.0Unranked · 2 rankable rows	86.5Unranked · 2 rankable rows	Not comparableProvisional lane · 0 vs 2 weighted rows	Not comparable
Math	27.7Unranked · 3 rankable rows	Not ranked	Not comparableProvisional lane · 1 vs 0 weighted rows	Not comparable
Multilingual	Not ranked	Not ranked	Not comparableProvisional lane · 0 vs 0 weighted rows	Not comparable
Multimodal	Not ranked	87.4#5/48	Not comparableProvisional lane · 0 vs 2 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
LFM2.5-8B-A1B: 56.5%Qwen3.8 Max: 82.8%Normalized gap 26.3LFM2.5-8B-A1B source Qwen3.8 Max 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

LFM2.5-8B-A1B: Self-hosted; infrastructure cost varies; Fits in one request
Qwen3.8 Max: API rate not published; Fits in one request

LFM2.5-8B-A1B has no comparable published API token rate. Qwen3.8 Max has no comparable published API token rate.

Repository review

50K fresh input + 3K output tokens

LFM2.5-8B-A1B: Self-hosted; infrastructure cost varies; Fits in one request
Qwen3.8 Max: API rate not published; Fits in one request

LFM2.5-8B-A1B has no comparable published API token rate. Qwen3.8 Max has no comparable published API token rate.

Cache-heavy agent loop

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

LFM2.5-8B-A1B: Self-hosted; infrastructure cost varies; Does not fit in one request; Cached-input rate unavailable
Qwen3.8 Max: API rate not published; Fits in one request; Cached-input rate unavailable

LFM2.5-8B-A1B does not fit this workload in one request. LFM2.5-8B-A1B has no comparable published API token rate. Qwen3.8 Max 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.

SpecificationLFM2.5-8B-A1BQwen3.8 Max

Context window

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

LFM2.5-8B-A1B

128K

Qwen3.8 Max

Alibaba Cloud Model Studio pricing

API model ID

LFM2.5-8B-A1B

Not sourced

Qwen3.8 Max

qwen3.8-max

Alibaba Cloud Model Studio pricing

Cached-input rate

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

LFM2.5-8B-A1B

No comparable hosted API rate

Qwen3.8 Max

No comparable hosted API rate

Alibaba Cloud Model Studio pricing

Documented inputs

LFM2.5-8B-A1B

Not sourced

Qwen3.8 Max

Not sourced

Documented outputs

LFM2.5-8B-A1B

Not sourced

Qwen3.8 Max

Not sourced

Provider availability

LFM2.5-8B-A1B

Not sourced

Qwen3.8 Max

Not sourced

Reasoning profile

LFM2.5-8B-A1B

Reasoning

Qwen3.8 Max

Reasoning

Weight access

LFM2.5-8B-A1B

Open Weight

Qwen3.8 Max

Open Weight

License

LFM2.5-8B-A1B

Open Weight

Qwen3.8 Max

Open Weight

Release date

LFM2.5-8B-A1B

2026-05-28

Qwen3.8 Max

2026-08-03

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: Qwen3.8 Max has the higher public score, 71.76 versus 39.15, and the 90% score intervals do not overlap.
Workload cost: A complete comparable API-rate estimate is not available for both models.
Context tradeoff: Qwen3.8 Max 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 evidence66 rows

Agentic

BFCL v4
LFM2.5-8B-A1B49.7%
Source
Qwen3.8 Max—
Not directly comparable
τ²-bench results
LFM2.5-8B-A1B88.1%
Source
Qwen3.8 Max—
Not directly comparable
Terminal-Bench 2.1
LFM2.5-8B-A1B—
Qwen3.8 Max86.6%
Source
Not directly comparable
CoWorkBench
LFM2.5-8B-A1B—
Qwen3.8 Max74.8%
Source
Not directly comparable
JobBench
LFM2.5-8B-A1B—
Qwen3.8 Max53.4%
Source
Not directly comparable
skillsBench
LFM2.5-8B-A1B—
Qwen3.8 Max70.2%
Source
Not directly comparable
Agents' Last Exam
LFM2.5-8B-A1B—
Qwen3.8 Max52.4%
Source
Not directly comparable
AutomationBench
LFM2.5-8B-A1B—
Qwen3.8 Max27.3%
Source
Not directly comparable
Toolathlon-Verified
LFM2.5-8B-A1B—
Qwen3.8 Max72.5%
Source
Not directly comparable
WideResearch
LFM2.5-8B-A1B—
Qwen3.8 Max81.9%
Source
Not directly comparable
HLE w/ tools
LFM2.5-8B-A1B—
Qwen3.8 Max56.2%
Source
Not directly comparable
OSWorld-Verified
LFM2.5-8B-A1B—
Qwen3.8 Max86.1%
Source
Not directly comparable
OSWorld 2.0
LFM2.5-8B-A1B—
Qwen3.8 Max19.4%
Source
Not directly comparable
WebArena-Verified
LFM2.5-8B-A1B—
Qwen3.8 Max66.8%
Source
Not directly comparable
AndroidWorld
LFM2.5-8B-A1B—
Qwen3.8 Max85.3%
Source
Not directly comparable
MobileWorld
LFM2.5-8B-A1B—
Qwen3.8 Max77.8%
Source
Not directly comparable
Terminal-Bench 2.1 (Vals)
LFM2.5-8B-A1B—
Qwen3.8 Max67.4%
Source
Not directly comparable

Coding

Terminal-Bench 2.1
LFM2.5-8B-A1B—
Qwen3.8 Max86.6%
Source
Not directly comparable
SWE-bench Pro
LFM2.5-8B-A1B—
Qwen3.8 Max67.7%
Source
Not directly comparable
DeepSWE
LFM2.5-8B-A1B—
Qwen3.8 Max56.6%
Source
Not directly comparable
NL2Repo
LFM2.5-8B-A1B—
Qwen3.8 Max55.9%
Source
Not directly comparable
FrontierSWE
LFM2.5-8B-A1B—
Qwen3.8 Max73.5%
Source
Not directly comparable
MLS-Bench Lite
LFM2.5-8B-A1B—
Qwen3.8 Max41.0%
Source
Not directly comparable
PaperBench
LFM2.5-8B-A1B—
Qwen3.8 Max93.0%
Source
Not directly comparable
VulcanBench v3
LFM2.5-8B-A1B—
Qwen3.8 Max81.2%
Source
Not directly comparable
OpenHarmony Bench
LFM2.5-8B-A1B—
Qwen3.8 Max60.8%
Source
Not directly comparable
FrontierSWE v2
LFM2.5-8B-A1B—
Qwen3.8 Max15.8%
Source
Not directly comparable
LiveCodeBench (Vals)
LFM2.5-8B-A1B—
Qwen3.8 Max87.9%
Source
Not directly comparable
SWE-bench (Vals)
LFM2.5-8B-A1B—
Qwen3.8 Max85.6%
Source
Not directly comparable

Reasoning

MRCRv2
LFM2.5-8B-A1B—
Qwen3.8 Max92.9%
Source
Not directly comparable
LongBench v2
LFM2.5-8B-A1B—
Qwen3.8 Max66.3%
Source
Not directly comparable

Knowledge

GPQA
LFM2.5-8B-A1B—
Qwen3.8 Max92.6%
Source
Not directly comparable
GPQA-D
LFM2.5-8B-A1B—
Qwen3.8 Max92.6%
Source
Not directly comparable
HLE
LFM2.5-8B-A1B—
Qwen3.8 Max43.6%
Source
Not directly comparable
HLE w/o tools
LFM2.5-8B-A1B—
Qwen3.8 Max43.6%
Source
Not directly comparable
GPQA Diamond (Vals)
LFM2.5-8B-A1B—
Qwen3.8 Max93.7%
Source
Not directly comparable
MMLU-Pro (Vals)
LFM2.5-8B-A1B—
Qwen3.8 Max88.6%
Source
Not directly comparable

Math

MATH-500
LFM2.5-8B-A1B88.8%
Source
Qwen3.8 Max—
Not directly comparable
AIME 2025
LFM2.5-8B-A1B42.5%
Source
Qwen3.8 Max—
Not directly comparable
AIME26
LFM2.5-8B-A1B50.0%
Source
Qwen3.8 Max—
Not directly comparable

Multimodal

MMMU-Pro
LFM2.5-8B-A1B—
Qwen3.8 Max82.3%
Source
Not directly comparable
MathVision
LFM2.5-8B-A1B—
Qwen3.8 Max95.2%
Source
Not directly comparable
MathVision w/ Python
LFM2.5-8B-A1B—
Qwen3.8 Max97.7%
Source
Not directly comparable
BabyVision
LFM2.5-8B-A1B—
Qwen3.8 Max82.0%
Source
Not directly comparable
BabyVision w/ Python
LFM2.5-8B-A1B—
Qwen3.8 Max91.3%
Source
Not directly comparable
ZeroBench
LFM2.5-8B-A1B—
Qwen3.8 Max24.0%
Source
Not directly comparable
ZeroBench w/ Python
LFM2.5-8B-A1B—
Qwen3.8 Max49.0%
Source
Not directly comparable
MedXpertQA (MM)
LFM2.5-8B-A1B—
Qwen3.8 Max80.4%
Source
Not directly comparable
ScreenSpot Pro
LFM2.5-8B-A1B—
Qwen3.8 Max84.5%
Source
Not directly comparable
Vision2Web
LFM2.5-8B-A1B—
Qwen3.8 Max69.0%
Source
Not directly comparable
CharXiv w/o tools
LFM2.5-8B-A1B—
Qwen3.8 Max88.4%
Source
Not directly comparable
CharXiv
LFM2.5-8B-A1B—
Qwen3.8 Max93.5%
Source
Not directly comparable
OmniDocBench 1.5
LFM2.5-8B-A1B—
Qwen3.8 Max92.1%
Source
Not directly comparable
OCRBench V2
LFM2.5-8B-A1B—
Qwen3.8 Max74.2%
Source
Not directly comparable
CC-OCR
LFM2.5-8B-A1B—
Qwen3.8 Max79.6%
Source
Not directly comparable
RealWorldQA
LFM2.5-8B-A1B—
Qwen3.8 Max88.0%
Source
Not directly comparable
ERQA
LFM2.5-8B-A1B—
Qwen3.8 Max77.8%
Source
Not directly comparable
SimpleVQA
LFM2.5-8B-A1B—
Qwen3.8 Max75.0%
Source
Not directly comparable
PerceptionBench
LFM2.5-8B-A1B—
Qwen3.8 Max63.5%
Source
Not directly comparable
Video-MME (with subtitle)
LFM2.5-8B-A1B—
Qwen3.8 Max90.4%
Source
Not directly comparable
VideoMMMU
LFM2.5-8B-A1B—
Qwen3.8 Max88.7%
Source
Not directly comparable
MMVU
LFM2.5-8B-A1B—
Qwen3.8 Max82.4%
Source
Not directly comparable
MLVU (M-Avg)
LFM2.5-8B-A1B—
Qwen3.8 Max90.8%
Source
Not directly comparable
LVBench
LFM2.5-8B-A1B—
Qwen3.8 Max81.8%
Source
Not directly comparable

Instruction following

IFEval
LFM2.5-8B-A1B91.8%
Source
Qwen3.8 Max—
Not directly comparable
IFBench
LFM2.5-8B-A1B56.5%
Source
Qwen3.8 Max82.8%
Source
Qwen3.8 Max leads this result

Questions

Which is better, LFM2.5-8B-A1B or Qwen3.8 Max?

Qwen3.8 Max has the higher public score, 71.76 versus 39.15, and the 90% score intervals do not overlap. That is the clearest overall quality signal in the current public data.

Which is better for coding, LFM2.5-8B-A1B or Qwen3.8 Max?

LFM2.5-8B-A1B is not ranked on the public lane for coding, so no winner is named for coding.

Which is better for agentic tasks, LFM2.5-8B-A1B or Qwen3.8 Max?

Qwen3.8 Max scores higher for agentic tasks on the public lane, 67.3 to 43.5. LFM2.5-8B-A1B is 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, LFM2.5-8B-A1B or Qwen3.8 Max?

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, LFM2.5-8B-A1B or Qwen3.8 Max?

Qwen3.8 Max has the larger documented context window: 1M, compared with 128K.

Related comparisons

Compare API pricing Read the methodology Open the model selector

Last updated September 15, 2026

Watch LFM2.5-8B-A1B vs Qwen3.8 Max

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

Read a sample issue

Join 2,000+ readers.

LFM2.5-8B-A1B vs Qwen3.8 Max

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

Knowledge

Instruction following

Agentic

Coding

Reasoning

Math

Multilingual

Multimodal

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

Multimodal

Instruction following

Questions

Which is better, LFM2.5-8B-A1B or Qwen3.8 Max?

Which is better for coding, LFM2.5-8B-A1B or Qwen3.8 Max?

Which is better for agentic tasks, LFM2.5-8B-A1B or Qwen3.8 Max?

Which costs less, LFM2.5-8B-A1B or Qwen3.8 Max?

Which has the larger context window, LFM2.5-8B-A1B or Qwen3.8 Max?

Related comparisons

Watch LFM2.5-8B-A1B vs Qwen3.8 Max