LLMBenchmark.ai

Enter opens the highlighted result. Esc closes.

Qwen3.8 27B vs Beam 1.0 501B A23B

Updated October 21, 2026. 26 benchmarks have a result for both. Coding is ranked on the Vibe Coding Index; every other board is shown as published.

Replace Qwen3.8 27B Replace Beam 1.0 501B A23B
Qwen3.8 27BQwen
Coding
73.6
Overall
33.7
Output $/M
$1.35
22
R Beam 1.0 501B A23BReflection
Coding
33.8
Overall
No data
Output $/M
$1.00
4
vs boards won of 26 shared

Decision reading

Pick Qwen3.8 27B for coding: it scores 73.6 on the Vibe Coding Index against 33.8 for Beam 1.0 501B A23B.

Pick Beam 1.0 501B A23B when output tokens cost 1.4x less, or when the context window is larger, 1.05M against 1M tokens.

Model AQwen3.8 27BQwen73.6/1002 independent benchmarks$1.26 per solved taskModel BBeam 1.0 501B A23BReflection33.8/1001 independent benchmarkNo cost per solved task
Shared results
26
Qwen3.8 27B only
59
Beam 1.0 501B A23B only
2
Like-for-like categories
5 of 9

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

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 73.6 from 2 benchmarks, Beam 1.0 501B A23B 33.8 from 1 benchmark.
Widest category
Coding, 15.1 points apart on average over 5 shared benchmarks.
Widest single gap
LMArena WebDev: React: Qwen3.8 27B 1602 against Beam 1.0 501B A23B 1443, 26.2 points apart on a 0 to 100 scale.
Output price
Qwen3.8 27B $1.35 against Beam 1.0 501B A23B $1.00 per million output tokens, 1.4x apart.
Context window
Qwen3.8 27B takes 1M tokens, Beam 1.0 501B A23B 1.05M.

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 27BBarsBeam 1.0 501B A23BTies
Coding50
Reasoning30
Math02
Knowledge40
General102
All shared boards224
CodingLike-for-likeFour of these boards feed the Vibe Coding Index50
Coding benchmarks for Qwen3.8 27B and Beam 1.0 501B A23B
BenchmarkQwen3.8 27BBeam 1.0 501B A23BLeader
LiveBench coding86.8%Not measuredOnly Qwen3.8 27B measured
LMArena WebDev15931437Qwen3.8 27B
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
LMArena Text: coding14831473Qwen3.8 27B
LMArena Text: software and IT14781461Qwen3.8 27B
LMArena WebDev: React16021443Qwen3.8 27B
LMArena WebDev: HTML15701436Qwen3.8 27B
LMArena WebDev: from an image1577Not measuredOnly Qwen3.8 27B measured
Neither model has a result on the other 20 boards here
IntelligenceDirectional onlyPublic boards, shown as published, kept out of the index2 boards, none shared
Intelligence benchmarks for Qwen3.8 27B and Beam 1.0 501B A23B
BenchmarkQwen3.8 27BBeam 1.0 501B A23BLeader
AA Intelligence Index33.7Not measuredOnly Qwen3.8 27B measured
Epoch Capabilities Index151.7Not measuredOnly Qwen3.8 27B measured
Agentic workDirectional onlyPublic boards, shown as published, kept out of the index10 boards, none shared
Agentic work benchmarks for Qwen3.8 27B and Beam 1.0 501B A23B
BenchmarkQwen3.8 27BBeam 1.0 501B A23BLeader
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
APEX-Agents47.5%Not measuredOnly Qwen3.8 27B measured
Neither model has a result on the other 29 boards here
ReasoningLike-for-likePublic boards, shown as published, kept out of the index30
Reasoning benchmarks for Qwen3.8 27B and Beam 1.0 501B A23B
BenchmarkQwen3.8 27BBeam 1.0 501B A23BLeader
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%Not measuredOnly Qwen3.8 27B measured
Humanity's Last Exam (Artificial Analysis run)33.9%Not measuredOnly Qwen3.8 27B measured
ARC-AGI87.5%Not measuredOnly Qwen3.8 27B measured
ARC-AGI-242.4%Not measuredOnly Qwen3.8 27B measured
LMArena Text: hard prompts14591442Qwen3.8 27B
LMArena Text: expert14791447Qwen3.8 27B
LMArena Text: hard prompts in English14611451Qwen3.8 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 9 boards here
MathLike-for-likePublic boards, shown as published, kept out of the index02
Math benchmarks for Qwen3.8 27B and Beam 1.0 501B A23B
BenchmarkQwen3.8 27BBeam 1.0 501B A23BLeader
ProofBench16.0%Not measuredOnly Qwen3.8 27B measured
LMArena Text: math14551459Beam 1.0 501B A23B
LMArena Text: mathematical work14581475Beam 1.0 501B A23B
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 Beam 1.0 501B A23B
BenchmarkQwen3.8 27BBeam 1.0 501B A23BLeader
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: science14671430Qwen3.8 27B
LMArena Text: medicine and healthcare14371376Qwen3.8 27B
LMArena Text: legal and government14471433Qwen3.8 27B
LMArena Text: business and finance14381430Qwen3.8 27B
Neither model has a result on the other 8 boards here
Web designDirectional onlyPublic boards, shown as published, kept out of the index3 boards, none shared
Web design benchmarks for Qwen3.8 27B and Beam 1.0 501B A23B
BenchmarkQwen3.8 27BBeam 1.0 501B A23BLeader
Design Arena: websitesNot measured1253Only Beam 1.0 501B A23B measured
Design Arena: game devNot measured1222Only Beam 1.0 501B A23B measured
LiveBench design75.9%Not measuredOnly Qwen3.8 27B measured
Neither model has a result on the other 6 boards here
GeneralLike-for-likePublic boards, shown as published, kept out of the index102
General benchmarks for Qwen3.8 27B and Beam 1.0 501B A23B
BenchmarkQwen3.8 27BBeam 1.0 501B A23BLeader
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
Long context reasoning (AA-LCR)82.0%Not measuredOnly Qwen3.8 27B measured
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 Text14401425Qwen3.8 27B
LMArena Text: instruction following14361430Qwen3.8 27B
LMArena Text: creative writing13831390Beam 1.0 501B A23B
LMArena Vision1269Not measuredOnly Qwen3.8 27B measured
LMArena Text: English14481436Qwen3.8 27B
LMArena Text: non-English14281413Qwen3.8 27B
LMArena Text: Chinese15011424Qwen3.8 27B
LMArena Text: German1443Not measuredOnly Qwen3.8 27B measured
LMArena Text: Japanese1387Not measuredOnly Qwen3.8 27B measured
LMArena Text: Korean1384Not measuredOnly Qwen3.8 27B measured
LMArena Text: French14651461Qwen3.8 27B
LMArena Text: Spanish1449Not measuredOnly Qwen3.8 27B measured
LMArena Text: Russian14151412Qwen3.8 27B
LMArena Text: Polish1439Not measuredOnly Qwen3.8 27B measured
LMArena Text: multi-turn14401414Qwen3.8 27B
LMArena Text: longer queries14481429Qwen3.8 27B
LMArena Text: writing and language14051400Qwen3.8 27B
LMArena Text: entertainment and media13931399Beam 1.0 501B A23B
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 16 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 Beam 1.0 501B A23B
BenchmarkQwen3.8 27BBeam 1.0 501B A23BLeader
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$0.0007Fits in one request
Beam 1.0 501B A23B$0.0008Fits in one request

Qwen3.8 27B costs less

Repository review

50K input + 3K output tokens

Qwen3.8 27B$0.0060Fits in one request
Beam 1.0 501B A23B$0.018Fits in one request

Qwen3.8 27B costs less

Agent loop

200K input + 10K output tokens

Qwen3.8 27B$0.021Fits in one request
Beam 1.0 501B A23B$0.070Fits 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 Beam 1.0 501B A23B
SpecificationQwen3.8 27BReflection Beam 1.0 501B A23B
MakerQwenReflection
WeightsOpen weightsOpen weights
ReleasedAugust 14, 2026October 16, 2026
Context window1M1.05M
Input $/Mtok$0.04$0.30
Output $/Mtok$1.35$1.00
Coding index73.633.8
Independent measurements behind it21
Cost per solved task$1.26Not computable
SourceOpenRouter listingOpenRouter listing

Questions

Is Qwen3.8 27B better than Beam 1.0 501B A23B for coding?

Pick Qwen3.8 27B for coding: it scores 73.6 on the Vibe Coding Index against 33.8 for Beam 1.0 501B A23B. Pick Beam 1.0 501B A23B when output tokens cost 1.4x less, or when the context window is larger, 1.05M against 1M tokens.

Which is cheaper, Qwen3.8 27B or Beam 1.0 501B A23B?

Qwen3.8 27B costs $0.04 per million input tokens and $1.35 per million output tokens; Beam 1.0 501B A23B costs $0.30 and $1.00. Cost per solved task is known for Qwen3.8 27B only.

Which has the bigger context window, Qwen3.8 27B or Beam 1.0 501B A23B?

Beam 1.0 501B A23B, with 1.05M tokens against 1M.

How many benchmarks is the Qwen3.8 27B vs Beam 1.0 501B A23B comparison based on?

26 benchmarks have a result for both models as of October 21, 2026. Qwen3.8 27B's coding index rests on 2 independent benchmarks and Beam 1.0 501B A23B's on 1. Where only one model has a result, the page says so instead of naming a winner.

Compare next