Enter opens the highlighted result. Esc closes.

Qwen3.8 27B vs Step 3.5 Flash

Qwen3.8 27B leads Step 3.5 Flash on the Artificial Analysis Intelligence Index, 33.7 to 17.0, and Step 3.5 Flash costs 8.5x less per output token ($0.30 vs $2.55 per million). Updated October 21, 2026; 32 benchmarks have a result for both.

Replace Qwen3.8 27B Replace Step 3.5 Flash
Qwen3.8 27BQwen
Coding
74.2
Overall
59.6
Output $/M
$2.55
32
Step 3.5 FlashStepFun
Coding
No data
Overall
No data
Output $/M
$0.30
0
vs boards won of 32 shared

Qwen3.8 27B or Step 3.5 Flash: which to pick

Only Qwen3.8 27B has a coding score, 74.2 on the Vibe Coding Index; Step 3.5 Flash has none, so the two cannot be ranked on coding.

Pick Step 3.5 Flash for output tokens 8.5x cheaper ($0.30 vs $2.55 per million).

Model AQwen3.8 27BQwen74.2/1004 independent benchmarks$1.26 per solved taskModel BStep 3.5 FlashStepFunNot benchmarkedNo coding benchmarkNo cost per solved task
Shared results
32
Qwen3.8 27B only
53
Step 3.5 Flash only
3
Like-for-like categories
6 of 9

Only benchmarks that tested both models count toward a pick; a result only one model has is shown but not compared.

Which one for your work

Each card is decided by one number both models carry. Where only one has it, the card says so and makes no call.

At a glance

Coding index
Qwen3.8 27B 74.2 from 4 benchmarks, Step 3.5 Flash not benchmarked.
Widest category
Intelligence, 16.7 points apart on average over 1 shared benchmark.
Widest single gap
Long context reasoning (AA-LCR): Qwen3.8 27B 82.0% against Step 3.5 Flash 63.0%, 19.0 points apart on a 0 to 100 scale.
Output price
Qwen3.8 27B $2.55 against Step 3.5 Flash $0.30 per million output tokens, 8.5x apart.
Context window
Qwen3.8 27B takes 1M tokens, Step 3.5 Flash 262K.

Benchmark by benchmark

First the score: boards each model leads, per category, where both have a result. Then every board, a category to a fold; bars are each result on a 0 to 100 scale, so compare the two in a row, not one row with another.

Boards won, where both have a result
CategoryQwen3.8 27BBarsStep 3.5 FlashTies
Coding20
Intelligence10
Reasoning50
Math20
Knowledge40
General180
All shared boards320
CodingLike-for-likeSix of these boards feed the Vibe Coding Index20
Coding benchmarks for Qwen3.8 27B and Step 3.5 Flash
BenchmarkQwen3.8 27BStep 3.5 FlashLeader
LiveBench coding86.8%Not measuredOnly Qwen3.8 27B measured
LMArena WebDev1593Not measuredOnly Qwen3.8 27B measured
AA Coding Index68.1Not measuredOnly Qwen3.8 27B measured
LiveCodeBench (Vals AI)84.0%Not measuredOnly Qwen3.8 27B measured
SWE-bench Verified (Vals AI)86.0%Not measuredOnly Qwen3.8 27B measured
Vibe Code Bench (Vals AI)64.8%Not measuredOnly Qwen3.8 27B measured
Code migration (Vals AI)14.2%Not measuredOnly Qwen3.8 27B measured
ProgramBench (Vals AI)0.0%Not measuredOnly Qwen3.8 27B measured
IOI (Vals AI)39.1%Not measuredOnly Qwen3.8 27B measured
Terminal-Bench 2.1 (Vals AI)58.4%Not measuredOnly Qwen3.8 27B measured
SkillsBench (Vals AI)38.1%Not measuredOnly Qwen3.8 27B measured
CyberBench (Vals AI)67.8%Not measuredOnly Qwen3.8 27B measured
SciCode46.6%Not measuredOnly Qwen3.8 27B measured
Terminal-Bench Hard (Artificial Analysis run)Not measured32.6%Only Step 3.5 Flash measured
LMArena Text: coding14831436Qwen3.8 27B
LMArena Text: software and IT14781434Qwen3.8 27B
LMArena WebDev: React1602Not measuredOnly Qwen3.8 27B measured
LMArena WebDev: HTML1570Not measuredOnly Qwen3.8 27B measured
LMArena WebDev: from an image1577Not measuredOnly Qwen3.8 27B measured
Neither model has a result on the other 20 boards here
IntelligenceLike-for-likePublic boards, shown as published, kept out of the index10
Intelligence benchmarks for Qwen3.8 27B and Step 3.5 Flash
BenchmarkQwen3.8 27BStep 3.5 FlashLeader
AA Intelligence Index33.717.0Qwen3.8 27B
Epoch Capabilities Index151.7Not measuredOnly Qwen3.8 27B measured
Agentic workDirectional onlyPublic boards, shown as published, kept out of the index11 boards, none shared
Agentic work benchmarks for Qwen3.8 27B and Step 3.5 Flash
BenchmarkQwen3.8 27BStep 3.5 FlashLeader
AA Agentic Index45.8Not measuredOnly Qwen3.8 27B measured
LiveBench research59.4%Not measuredOnly Qwen3.8 27B measured
LiveBench browser32.0%Not measuredOnly Qwen3.8 27B measured
LiveBench long agentic27.9%Not measuredOnly Qwen3.8 27B measured
Finance Agent (Vals AI)48.5%Not measuredOnly Qwen3.8 27B measured
Tax Agent Bench (Vals AI)30.9%Not measuredOnly Qwen3.8 27B measured
Legal Research Bench (Vals AI)36.1%Not measuredOnly Qwen3.8 27B measured
Excel Modeling Benchmark (Vals AI)59.7%Not measuredOnly Qwen3.8 27B measured
Terminal-Bench-Science (Vals AI)0.0%Not measuredOnly Qwen3.8 27B measured
τ²-Bench TelecomNot measured94.4%Only Step 3.5 Flash measured
APEX-Agents47.5%Not measuredOnly Qwen3.8 27B measured
Neither model has a result on the other 30 boards here
ReasoningLike-for-likePublic boards, shown as published, kept out of the index50
Reasoning benchmarks for Qwen3.8 27B and Step 3.5 Flash
BenchmarkQwen3.8 27BStep 3.5 FlashLeader
GPQA Diamond88.9%Not measuredOnly Qwen3.8 27B measured
LiveBench reasoning80.0%Not measuredOnly Qwen3.8 27B measured
MMMU Pro (Vals AI)83.9%Not measuredOnly Qwen3.8 27B measured
GPQA Diamond (Artificial Analysis run)90.5%83.1%Qwen3.8 27B
Humanity's Last Exam (Artificial Analysis run)33.9%24.5%Qwen3.8 27B
ARC-AGI87.5%Not measuredOnly Qwen3.8 27B measured
ARC-AGI-242.4%Not measuredOnly Qwen3.8 27B measured
LMArena Text: hard prompts14591411Qwen3.8 27B
LMArena Text: expert14791421Qwen3.8 27B
LMArena Text: hard prompts in English14611421Qwen3.8 27B
LMArena Vision: diagrams1285Not measuredOnly Qwen3.8 27B measured
LMArena Vision: homework1295Not measuredOnly Qwen3.8 27B measured
Neither model has a result on the other 10 boards here
MathLike-for-likePublic boards, shown as published, kept out of the index20
Math benchmarks for Qwen3.8 27B and Step 3.5 Flash
BenchmarkQwen3.8 27BStep 3.5 FlashLeader
ProofBench16.0%Not measuredOnly Qwen3.8 27B measured
LMArena Text: math14551408Qwen3.8 27B
LMArena Text: mathematical work14581421Qwen3.8 27B
Neither model has a result on the other 10 boards here
KnowledgeLike-for-likePublic boards, shown as published, kept out of the index40
Knowledge benchmarks for Qwen3.8 27B and Step 3.5 Flash
BenchmarkQwen3.8 27BStep 3.5 FlashLeader
MMLU Pro84.3%Not measuredOnly Qwen3.8 27B measured
LegalBench (Vals AI)82.4%Not measuredOnly Qwen3.8 27B measured
TaxEval (Vals AI)70.8%Not measuredOnly Qwen3.8 27B measured
MortgageTax (Vals AI)64.9%Not measuredOnly Qwen3.8 27B measured
MedCode (Vals AI)28.7%Not measuredOnly Qwen3.8 27B measured
LMArena Text: science14671415Qwen3.8 27B
LMArena Text: medicine and healthcare14371408Qwen3.8 27B
LMArena Text: legal and government14471408Qwen3.8 27B
LMArena Text: business and finance14381394Qwen3.8 27B
Neither model has a result on the other 9 boards here
Web designDirectional onlyPublic boards, shown as published, kept out of the index1 board, none shared
Web design benchmarks for Qwen3.8 27B and Step 3.5 Flash
BenchmarkQwen3.8 27BStep 3.5 FlashLeader
LiveBench design75.9%Not measuredOnly Qwen3.8 27B measured
Neither model has a result on the other 8 boards here
GeneralLike-for-likePublic boards, shown as published, kept out of the index180
General benchmarks for Qwen3.8 27B and Step 3.5 Flash
BenchmarkQwen3.8 27BStep 3.5 FlashLeader
LiveBench data analysis76.6%Not measuredOnly Qwen3.8 27B measured
SAGE (Vals AI)52.4%Not measuredOnly Qwen3.8 27B measured
MedScribe (Vals AI)83.8%Not measuredOnly Qwen3.8 27B measured
IFBenchNot measured66.5%Only Step 3.5 Flash measured
Long context reasoning (AA-LCR)82.0%63.0%Qwen3.8 27B
DTBench88.0%Not measuredOnly Qwen3.8 27B measured
LMCA41.4%Not measuredOnly Qwen3.8 27B measured
Surface Evolver Bench45.0%Not measuredOnly Qwen3.8 27B measured
LMArena Text14401403Qwen3.8 27B
LMArena Text: instruction following14361385Qwen3.8 27B
LMArena Text: creative writing13831357Qwen3.8 27B
LMArena Vision1269Not measuredOnly Qwen3.8 27B measured
LMArena Text: English14481417Qwen3.8 27B
LMArena Text: non-English14281385Qwen3.8 27B
LMArena Text: Chinese15011446Qwen3.8 27B
LMArena Text: German14431405Qwen3.8 27B
LMArena Text: Japanese13871353Qwen3.8 27B
LMArena Text: Korean13841351Qwen3.8 27B
LMArena Text: French14651421Qwen3.8 27B
LMArena Text: Spanish14491419Qwen3.8 27B
LMArena Text: Russian14151385Qwen3.8 27B
LMArena Text: Polish14391389Qwen3.8 27B
LMArena Text: multi-turn14401404Qwen3.8 27B
LMArena Text: longer queries14481402Qwen3.8 27B
LMArena Text: writing and language14051372Qwen3.8 27B
LMArena Text: entertainment and media13931360Qwen3.8 27B
LMArena Vision: English1267Not measuredOnly Qwen3.8 27B measured
LMArena Vision: Chinese1326Not measuredOnly Qwen3.8 27B measured
LMArena Vision: reading text1281Not measuredOnly Qwen3.8 27B measured
LMArena Vision: humor1235Not measuredOnly Qwen3.8 27B measured
LMArena Vision: creative writing1237Not measuredOnly Qwen3.8 27B measured
Neither model has a result on the other 15 boards here
Image generationNot comparablePublic boards, shown as published, kept out of the index0 boards, none shared
Image generation benchmarks for Qwen3.8 27B and Step 3.5 Flash
BenchmarkQwen3.8 27BStep 3.5 FlashLeader
Neither model has a result on 5 boards here

What each workload costs

Three fixed token mixes at each model's listed price, so per-token rates become a bill.

Chat turn

1K input + 500 output tokens

Qwen3.8 27B$0.0017Fits in one request
Step 3.5 Flash$0.0003Fits in one request

Step 3.5 Flash costs less

Repository review

50K input + 3K output tokens

Qwen3.8 27B$0.029Fits in one request
Step 3.5 Flash$0.0059Fits in one request

Step 3.5 Flash costs less

Agent loop

200K input + 10K output tokens

Qwen3.8 27B$0.111Fits in one request
Step 3.5 Flash$0.023Fits in one request

Step 3.5 Flash costs less

Listed OpenRouter rates, without cache or batch discounts. A workload that does not fit a model's context window would need splitting.

Specification differences

The leading value in a row is marked; a row with a missing value names no leader. Our two rows are marked in red.

Specifications of Qwen3.8 27B and Step 3.5 Flash
SpecificationQwen3.8 27BStepFun Step 3.5 Flash
MakerQwenStepFun
WeightsOpen weightsOpen weights
ReleasedAugust 14, 2026January 29, 2026
Context window1M262K
Input $/Mtok$0.425$0.10
Output $/Mtok$2.55$0.30
Coding index74.2Not benchmarked
Independent measurements behind it40
Cost per solved task$1.26Not computable
SourceOpenRouter listingOpenRouter listing

Questions

Is Qwen3.8 27B better than Step 3.5 Flash for coding?

Only Qwen3.8 27B has a coding score, 74.2 on the Vibe Coding Index; Step 3.5 Flash has none, so the two cannot be ranked on coding. Pick Step 3.5 Flash for output tokens 8.5x cheaper ($0.30 vs $2.55 per million).

Which is cheaper, Qwen3.8 27B or Step 3.5 Flash?

Qwen3.8 27B costs $0.425 per million input tokens and $2.55 per million output tokens; Step 3.5 Flash costs $0.10 and $0.30. Cost per solved task is known for Qwen3.8 27B only.

Which has the bigger context window, Qwen3.8 27B or Step 3.5 Flash?

Qwen3.8 27B, with 1M tokens against 262K.

How many benchmarks is the Qwen3.8 27B vs Step 3.5 Flash comparison based on?

32 benchmarks have a result for both models as of October 21, 2026. Qwen3.8 27B's coding index rests on 4 independent benchmarks and Step 3.5 Flash's on 0. Where only one model has a result, the page says so instead of naming a winner.

Compare next