Enter opens the highlighted result. Esc closes.

Qwen3.8 27B vs Inkling

Qwen3.8 27B leads Inkling on coding, 74.2 to 31.5 on the Vibe Coding Index, and Qwen3.8 27B costs 1.6x less per output token ($2.55 vs $4.05 per million). Updated October 22, 2026; 74 benchmarks have a result for both.

Replace Qwen3.8 27B Replace Inkling
Qwen3.8 27BQwen
Coding
74.2
Overall
59.6
Output $/M
$2.55
50
T InklingThinking Machines
Coding
31.5
Overall
46.5
Output $/M
$4.05
23
vs boards won of 74 shared, 1 tied

Qwen3.8 27B or Inkling: which to pick

Pick Qwen3.8 27B for coding: it scores 74.2 on the Vibe Coding Index against 31.5 for Inkling, though on 4 benchmarks against 6.

Pick Inkling for a lower cost per solved task ($0.76 against $1.26) and work measured by LiveCodeBench (Vals AI) or SciCode, where it leads.

Model AQwen3.8 27BQwen74.2/1004 independent benchmarks$1.26 per solved taskModel BInklingThinking Machines31.5/1006 independent benchmarks$0.76 per solved task
Shared results
74
Qwen3.8 27B only
11
Inkling only
28
Like-for-like categories
8 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, Inkling 31.5 from 6 benchmarks.
Widest category
Web design, 17.6 points apart on average over 1 shared benchmark.
Widest single gap
Vibe Code Bench (Vals AI): Qwen3.8 27B 64.8% against Inkling 19.2%, 45.6 points apart on a 0 to 100 scale.
Output price
Qwen3.8 27B $2.55 against Inkling $4.05 per million output tokens, 1.6x apart.
Context window
Qwen3.8 27B takes 1M tokens, Inkling 524K.

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 27BBarsInklingTies
Coding1421 tie
Intelligence20
Agentic work81
Reasoning100
Math12
Knowledge27
Web design10
General1211
All shared boards50231 tie
CodingLike-for-likeSix of these boards feed the Vibe Coding Index142
Coding benchmarks for Qwen3.8 27B and Inkling
BenchmarkQwen3.8 27BInklingLeader
LiveBench coding86.8%64.0%Qwen3.8 27B
LMArena WebDev15931414Qwen3.8 27B
AA Coding Index68.152.1Qwen3.8 27B
LiveCodeBench (Vals AI)84.0%85.5%Inkling
SWE-bench Verified (Vals AI)86.0%77.6%Qwen3.8 27B
Terminal-Bench 4 (Vals AI)Not measured0.5%Only Inkling measured
Vibe Code Bench (Vals AI)64.8%19.2%Qwen3.8 27B
Code migration (Vals AI)14.2%11.8%Qwen3.8 27B
ProgramBench (Vals AI)0.0%0.0%Tie
IOI (Vals AI)39.1%14.9%Qwen3.8 27B
Terminal-Bench 2.1 (Vals AI)58.4%47.6%Qwen3.8 27B
Vibe Code Bench 1-100 (Vals AI)Not measured7.3%Only Inkling measured
SkillsBench (Vals AI)38.1%26.1%Qwen3.8 27B
CyberBench (Vals AI)67.8%66.8%Qwen3.8 27B
SciCode46.6%47.0%Inkling
FrontierCodeNot measured14.0%Only Inkling measured
FrontierSWENot measured4.1%Only Inkling measured
WeirdMLNot measured32.3%Only Inkling measured
ALE-BenchNot measured946.0Only Inkling measured
LMArena Text: coding14831466Qwen3.8 27B
LMArena Text: software and IT14781465Qwen3.8 27B
LMArena WebDev: React16021404Qwen3.8 27B
LMArena WebDev: HTML15701399Qwen3.8 27B
LMArena WebDev: from an image1577Not measuredOnly Qwen3.8 27B measured
Neither model has a result on the other 15 boards here
IntelligenceLike-for-likePublic boards, shown as published, kept out of the index20
Intelligence benchmarks for Qwen3.8 27B and Inkling
BenchmarkQwen3.8 27BInklingLeader
AA Intelligence Index33.725.0Qwen3.8 27B
Epoch Capabilities Index151.7148.4Qwen3.8 27B
Agentic workLike-for-likePublic boards, shown as published, kept out of the index81
Agentic work benchmarks for Qwen3.8 27B and Inkling
BenchmarkQwen3.8 27BInklingLeader
AA Agentic Index45.822.5Qwen3.8 27B
Design Arena: agents building web appsNot measured1196Only Inkling measured
LiveBench research59.4%59.9%Inkling
LiveBench browser32.0%16.0%Qwen3.8 27B
LiveBench long agentic27.9%14.1%Qwen3.8 27B
SREBench (Vals AI)Not measured0.0%Only Inkling measured
Finance Agent (Vals AI)48.5%46.6%Qwen3.8 27B
Tax Agent Bench (Vals AI)30.9%15.3%Qwen3.8 27B
Legal Research Bench (Vals AI)36.1%28.4%Qwen3.8 27B
Harvey's Legal Agent Benchmark (Vals AI)Not measured3.8%Only Inkling measured
Excel Modeling Benchmark (Vals AI)59.7%38.4%Qwen3.8 27B
Terminal-Bench-Science (Vals AI)0.0%Not measuredOnly Qwen3.8 27B measured
APEX-Agents47.5%33.8%Qwen3.8 27B
Neither model has a result on the other 28 boards here
ReasoningLike-for-likePublic boards, shown as published, kept out of the index100
Reasoning benchmarks for Qwen3.8 27B and Inkling
BenchmarkQwen3.8 27BInklingLeader
GPQA Diamond88.9%87.1%Qwen3.8 27B
LiveBench reasoning80.0%78.3%Qwen3.8 27B
MMMU Pro (Vals AI)83.9%78.5%Qwen3.8 27B
GPQA Diamond (Artificial Analysis run)90.5%87.2%Qwen3.8 27B
Humanity's Last Exam (Artificial Analysis run)33.9%31.9%Qwen3.8 27B
GPQA Diamond (Epoch AI run)Not measured88.3%Only Inkling measured
ARC-AGI87.5%79.5%Qwen3.8 27B
ARC-AGI-242.4%36.5%Qwen3.8 27B
SimpleBenchNot measured50.0%Only Inkling measured
Chess puzzlesNot measured21.0%Only Inkling measured
LMArena Text: hard prompts14591451Qwen3.8 27B
LMArena Text: expert14791466Qwen3.8 27B
LMArena Text: hard prompts in English14611448Qwen3.8 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 7 boards here
MathLike-for-likePublic boards, shown as published, kept out of the index12
Math benchmarks for Qwen3.8 27B and Inkling
BenchmarkQwen3.8 27BInklingLeader
OTIS Mock AIMENot measured88.9%Only Inkling measured
FrontierMath, tiers 1 to 3Not measured33.3%Only Inkling measured
FrontierMath, tier 4Not measured4.9%Only Inkling measured
ProofBench16.0%0.0%Qwen3.8 27B
LMArena Text: math14551479Inkling
LMArena Text: mathematical work14581460Inkling
Neither model has a result on the other 7 boards here
KnowledgeLike-for-likePublic boards, shown as published, kept out of the index27
Knowledge benchmarks for Qwen3.8 27B and Inkling
BenchmarkQwen3.8 27BInklingLeader
MMLU Pro84.3%86.3%Inkling
LegalBench (Vals AI)82.4%83.0%Inkling
CorpFin (Vals AI)Not measured68.6%Only Inkling measured
TaxEval (Vals AI)70.8%75.3%Inkling
MortgageTax (Vals AI)64.9%64.0%Qwen3.8 27B
MedCode (Vals AI)28.7%41.2%Inkling
SimpleQA VerifiedNot measured40.3%Only Inkling measured
LMArena Text: science14671455Qwen3.8 27B
LMArena Text: medicine and healthcare14371447Inkling
LMArena Text: legal and government14471448Inkling
LMArena Text: business and finance14381438Inkling
Neither model has a result on the other 7 boards here
Web designLike-for-likePublic boards, shown as published, kept out of the index10
Web design benchmarks for Qwen3.8 27B and Inkling
BenchmarkQwen3.8 27BInklingLeader
Design Arena: websitesNot measured1224Only Inkling measured
Design Arena: UI componentsNot measured1198Only Inkling measured
Design Arena: game devNot measured1165Only Inkling measured
Design Arena: data vizNot measured1156Only Inkling measured
Design Arena: 3DNot measured1160Only Inkling measured
Design Arena: all code categoriesNot measured1205Only Inkling measured
Design Arena: SVGNot measured1108Only Inkling measured
Design Arena: ASCII artNot measured1089Only Inkling measured
LiveBench design75.9%58.3%Qwen3.8 27B
GeneralLike-for-likePublic boards, shown as published, kept out of the index1211
General benchmarks for Qwen3.8 27B and Inkling
BenchmarkQwen3.8 27BInklingLeader
Vals IndexNot measured29.0%Only Inkling measured
LiveBench data analysis76.6%72.8%Qwen3.8 27B
Public Benefits Bench (Vals AI)Not measured58.5%Only Inkling measured
SAGE (Vals AI)52.4%36.5%Qwen3.8 27B
MedScribe (Vals AI)83.8%85.4%Inkling
Vals Multimodal IndexNot measured49.4%Only Inkling measured
Long context reasoning (AA-LCR)82.0%77.3%Qwen3.8 27B
DTBench88.0%87.5%Qwen3.8 27B
LMCA41.4%37.6%Qwen3.8 27B
Surface Evolver Bench45.0%Not measuredOnly Qwen3.8 27B measured
LMArena Text14401442Inkling
LMArena Text: instruction following14361425Qwen3.8 27B
LMArena Text: creative writing13831390Inkling
LMArena Vision1269Not measuredOnly Qwen3.8 27B measured
LMArena Text: English14481442Qwen3.8 27B
LMArena Text: non-English14281434Inkling
LMArena Text: Chinese15011493Qwen3.8 27B
LMArena Text: German14431445Inkling
LMArena Text: Japanese13871429Inkling
LMArena Text: Korean13841403Inkling
LMArena Text: French14651455Qwen3.8 27B
LMArena Text: Spanish14491449Inkling
LMArena Text: Russian14151430Inkling
LMArena Text: Polish14391450Inkling
LMArena Text: multi-turn14401437Qwen3.8 27B
LMArena Text: longer queries14481435Qwen3.8 27B
LMArena Text: writing and language14051404Qwen3.8 27B
LMArena Text: entertainment and media13931394Inkling
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 13 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 Inkling
BenchmarkQwen3.8 27BInklingLeader
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
Inkling$0.0030Fits in one request

Qwen3.8 27B costs less

Repository review

50K input + 3K output tokens

Qwen3.8 27B$0.029Fits in one request
Inkling$0.061Fits in one request

Qwen3.8 27B costs less

Agent loop

200K input + 10K output tokens

Qwen3.8 27B$0.111Fits in one request
Inkling$0.235Fits in one request

Qwen3.8 27B 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 Inkling
SpecificationQwen3.8 27BThinking Machines Inkling
MakerQwenThinking Machines
WeightsOpen weightsOpen weights
ReleasedAugust 14, 2026July 17, 2026
Context window1M524K
Input $/Mtok$0.425$0.975
Output $/Mtok$2.55$4.05
Coding index74.231.5
Independent measurements behind it46
Cost per solved task$1.26$0.76
SourceOpenRouter listingOpenRouter listing

Questions

Is Qwen3.8 27B better than Inkling for coding?

Pick Qwen3.8 27B for coding: it scores 74.2 on the Vibe Coding Index against 31.5 for Inkling, though on 4 benchmarks against 6. Pick Inkling for a lower cost per solved task ($0.76 against $1.26) and work measured by LiveCodeBench (Vals AI) or SciCode, where it leads.

Which is cheaper, Qwen3.8 27B or Inkling?

Qwen3.8 27B costs $0.425 per million input tokens and $2.55 per million output tokens; Inkling costs $0.975 and $4.05. Per solved SWE-bench Verified task, Qwen3.8 27B costs $1.26 and Inkling $0.76.

Which has the bigger context window, Qwen3.8 27B or Inkling?

Qwen3.8 27B, with 1M tokens against 524K.

How many benchmarks is the Qwen3.8 27B vs Inkling comparison based on?

74 benchmarks have a result for both models as of October 22, 2026. Qwen3.8 27B's coding index rests on 4 independent benchmarks and Inkling's on 6. Where only one model has a result, the page says so instead of naming a winner.

Compare next

Embed this comparison on your site

Free under CC BY 4.0, while the source link stays visible. Preview the frame · Terms and other embeds

<iframe src="https://llmbenchmark.ai/embed/compare/qwen3-8-27b-vs-thinkingmachines-inkling" title="Qwen3.8 27B vs Inkling, LLMBenchmark.ai" width="100%" height="360" style="border:0;max-width:100%" loading="lazy"></iframe>
<p style="font:13px/1.4 sans-serif;margin:6px 0 0">Source: <a href="https://llmbenchmark.ai/compare/qwen3-8-27b-vs-thinkingmachines-inkling">LLMBenchmark.ai, Qwen3.8 27B vs Inkling</a></p>