RankingEffort benchmarks

Qwen3.8-27B · Medium

Compare models

77 measured results · 6 effort settings · 24 benchmark/harness combinations

Model evidence and sources →

Change model

Model

Current configurationSources

Compare with

Choose a second model and its effort to see both profiles and benchmark differences below.

Compare withModel and configuration

Capability profile

One configuration, seven capabilities

Comparison score · 0–100

AQwen3.8-27B · Medium2 of 7 capabilities supported

Gaps are unknown, not zero. Hollow points are preliminary.

Agentic
A43.8Supported
Hard reasoning
A15.4Supported
Coding
A36.1Preliminary
Human pref
Unknown
Knowledge
A5.7Preliminary
Multimodal
A20.0Preliminary
Long context
A50.7Preliminary

Each profile belongs to the named configuration. Capability scores use a shared panel for all axes, distinct from the overall ranking panel; they are not accuracy percentages or direct head-to-head wins.

Capability score evidence

Qwen3.8-27B · Medium: 13.1 overall comparison score. Meets the overall evidence requirements.

6 areas measured. The available task mix affects this estimate; missing areas remain unknown.

Shared-result disagreement. This aggregate order conflicts with matched benchmark results against 1 other configurations. Different test coverage and opponents affect the fit. Inspect published results before treating the rank as conclusive.

Which sources carry the weight?

ARC Prize · 55.2% direct comparison weight · independent of the model provider
Artificial Analysis · 44.8% direct comparison weight · independent of the model provider

Which benchmark families carry the weight?

ARC-AGI · 55.2% direct comparison weight
GPQA · 4.3% direct comparison weight
Humanity’s Last Exam · 4.3% direct comparison weight
Critpt · 4.3% direct comparison weight
omniscience · 4.2% direct comparison weight
Other families (9) · 27.6% direct comparison weight

These shares describe direct comparison weight, not fractions of the final score or confidence. Source balancing does not give every family equal weight or remove correlations between overlapping sources.

How sensitive is this score?

The audit removes one source at a time and refits against the same overall reference panel. These published results use the matching dataset audited 2026-10-23.

Both ranks use the same surviving configurations for each removal. The cohort starts with 444 configurations supported in the original audit, including separate effort settings. These are diagnostic positions, not the default leaderboard ranks. Eligibility is not rechecked after removal.

What changes when a source is removed?
Removed source	Rank before → after	Rank movement	Cohort remaining / lost support	Score after (change)
Artificial Analysis	138 → 134	Up 4	203 remaining / 241 lost	12.7 (-0.4 points)
ARC Prize	172 → 174	Down 2	444 remaining / 0 lost	16.0 (+2.9 points)
Datacurve	172 → 170	Up 2	444 remaining / 0 lost	10.4 (-2.7 points)
LiveCodeBench	172 → 169	Up 3	444 remaining / 0 lost	13.3 (+0.1 points)
Scale	172 → 170	Up 2	444 remaining / 0 lost	14.9 (+1.7 points)
Terminal-Bench	172 → 172	No movement	444 remaining / 0 lost	13.0 (-0.2 points)
Cognition	172 → 173	Down 1	444 remaining / 0 lost	12.5 (-0.7 points)
LiveBench	171 → 169	Up 2	443 remaining / 1 lost	13.5 (+0.4 points)
SWE-rebench	172 → 172	No movement	444 remaining / 0 lost	13.2 (+0.1 points)
Epoch AI	168 → 172	Down 4	431 remaining / 13 lost	13.1 (-0.0 points)
Vals AI	172 → 173	Down 1	443 remaining / 1 lost	12.5 (-0.7 points)
Provider reports · Anthropic	172 → 172	No movement	444 remaining / 0 lost	13.1 (-0.0 points)
Provider reports · Moonshot AI	172 → 170	Up 2	444 remaining / 0 lost	12.8 (-0.3 points)
Provider reports · Meta	172 → 172	No movement	444 remaining / 0 lost	13.1 (-0.0 points)
Provider reports · Mistral	172 → 172	No movement	444 remaining / 0 lost	13.1 (+0.0 points)
Provider reports · OpenAI	172 → 172	No movement	444 remaining / 0 lost	13.2 (+0.1 points)
All provider reports	172 → 170	Up 2	444 remaining / 0 lost	12.9 (-0.2 points)

Lost support means a configuration no longer has a comparison path to the complete reference panel. Its rank is unavailable. A surviving path alone does not establish leaderboard eligibility.

Overlapping sources may remain correlated even with source balancing. Leave-one-source-out movement is a sensitivity diagnostic, not a spread of source-specific ranks or a confidence measure. Scoring methodology · Dated configuration audit

Every benchmark family

Removed evidence	Recomputed score	Change
deepswe	10.4	-2.7
arc-agi	15.3	+2.2
critpt	15.0	+1.9
hle	14.6	+1.5
omniscience	14.0	+0.9
mmmu-pro	14.0	+0.9
arc-agi-3	12.4	-0.8
briefcase	12.4	-0.7
epoch-game-puzzles	12.4	-0.7
tau-bench	12.5	-0.7
frontiercode	12.5	-0.7
aa-lcr	12.5	-0.6
gdpval	12.6	-0.5
simpleqa	13.7	+0.5
automationbench	12.7	-0.4
terminal-bench	12.8	-0.3
mmlu-pro	13.4	+0.3
ifbench	12.8	-0.3
livebench-instructions	13.4	+0.3
multi-swe-bench	12.9	-0.2
gpqa	12.9	-0.2
livebench-language	13.3	+0.2
terminal-bench-science	13.3	+0.2
harvey	13.0	-0.2
aime	13.0	-0.1
enigma-eval	13.0	-0.1
apex-agents	13.3	+0.1
swe-atlas-qna	13.0	-0.1
livebench-coding	13.0	-0.1
swe-rebench	13.2	+0.1
gdp-pdf	13.0	-0.1
analyst-agent	13.2	+0.1
livecodebench	13.2	+0.1
vibe-code	13.0	-0.1
livebench-reasoning	13.1	-0.1
frontiermath	13.2	+0.1
swe-atlas-test-writing	13.1	-0.1
vals-code-migration	13.1	-0.0
vals-finance-agent	13.1	-0.0
enterprise-ops	13.2	+0.0
vals-legal-research	13.1	-0.0
itbench	13.1	-0.0
vals-tax-agent	13.1	-0.0
vals-excel-modeling	13.1	-0.0
math500	13.1	-0.0
livebench-math	13.1	+0.0
livebench-data	13.1	-0.0
swe-atlas-refactoring	13.1	-0.0
gmmlu	13.1	-0.0
swe-bench-pro	13.1	+0.0

What evidence is missing from the fit?

Across all collected settings for this model: 87 observations, 62 contributing, 0 matched without graph weight and 25 excluded. Counts do not establish rank eligibility.

5 · Benchmark family or source protocol has not been reviewed
4 · Original SciCode grading defects are under review; retained as supporting evidence, excluded from aggregate and capability scores. Corrected SciCode-Verified requires a separate protocol review. Sources: https://arxiv.org/abs/2608.04975 and https://artificialanalysis.ai/evaluations/scicode
4 · Evaluator has not been reviewed for independent admission
2 · Effort is not identified
7 · Supporting evidence outside the reviewed capability core
3 · No reviewed effort for this catalog observation; any separately collected effort measurement is counted separately

Inspect every observation and exclusion →

Benchmark results

These are published benchmark scores. Supported effort estimates also appear on the Capability leaderboard. Results from different harnesses occupy different rows. Missing cells stay unknown; no scores are copied between effort levels.

Benchmark / source	Low	Medium	XHigh	None	Non-reasoning	Unspecified
AA-Briefcase EloArtificial Analysis · Artificial Analysis published evaluation; see source for benchmark-specific settings Benchmark details AA-Briefcase v1.1 is an agentic knowledge work benchmark developed by Artificial Analysis. AA-Briefcase Elo is a combined metric that aggregates rubric pass rate, analytical quality Elo and presentation Elo · Higher is better · Evaluation results measured independently by Artificial Analysis Unit: elo. Source collected 2026-10-18.	1,271.57 Elo Reported settings Qwen3.8 27B (Low); Model source: https://artificialanalysis.ai/models/qwen3-8-27b-low	1,375.49 Elo Reported settings Qwen3.8 27B (Medium); Model source: https://artificialanalysis.ai/models/qwen3-8-27b-medium	1,395.27 Elo Reported settings Qwen3.8 27B (Xhigh); Model source: https://artificialanalysis.ai/models/qwen3-8-27b	Not reported	897.66 Elo Reported settings Qwen3.8 27B (Non-reasoning); Model source: https://artificialanalysis.ai/models/qwen3-8-27b-non-reasoning	Not reported
AA-LCR v1.1Artificial Analysis · Artificial Analysis published evaluation; see source for benchmark-specific settings Benchmark details Independently benchmarked by Artificial Analysis · Evaluation results measured independently by Artificial Analysis Unit: percent. Source collected 2026-10-18.	77.33% Reported settings Qwen3.8 27B (Low); Model source: https://artificialanalysis.ai/models/qwen3-8-27b-low	79.67% Reported settings Qwen3.8 27B (Medium); Model source: https://artificialanalysis.ai/models/qwen3-8-27b-medium	82% Reported settings Qwen3.8 27B (Xhigh); Model source: https://artificialanalysis.ai/models/qwen3-8-27b	Not reported	69.33% Reported settings Qwen3.8 27B (Non-reasoning); Model source: https://artificialanalysis.ai/models/qwen3-8-27b-non-reasoning	Not reported
AutomationBench-AAArtificial Analysis · Artificial Analysis published evaluation; see source for benchmark-specific settings Benchmark details Share of task objectives completed with no guardrail violations · Higher is better · Benchmark developed by Zapier · Independently benchmarked by Artificial Analysis · Evaluation results measured independently by Artificial Analysis Unit: percent. Source collected 2026-10-18.	30.1% Reported settings Qwen3.8 27B (Low); Model source: https://artificialanalysis.ai/models/qwen3-8-27b-low	40.25% Reported settings Qwen3.8 27B (Medium); Model source: https://artificialanalysis.ai/models/qwen3-8-27b-medium	48.24% Reported settings Qwen3.8 27B (Xhigh); Model source: https://artificialanalysis.ai/models/qwen3-8-27b	Not reported	12.43% Reported settings Qwen3.8 27B (Non-reasoning); Model source: https://artificialanalysis.ai/models/qwen3-8-27b-non-reasoning	Not reported
CritPtArtificial Analysis · Artificial Analysis published evaluation; see source for benchmark-specific settings Benchmark details Benchmark developed by Argonne and UIUC, with contributions from 60+ researchers globally. CritPt is under review by its authors following external feedback · Independently benchmarked by Artificial Analysis · Evaluation results measured independently by Artificial Analysis Unit: percent. Source collected 2026-10-18.	0% Reported settings Qwen3.8 27B (Low); Model source: https://artificialanalysis.ai/models/qwen3-8-27b-low	0% Reported settings Qwen3.8 27B (Medium); Model source: https://artificialanalysis.ai/models/qwen3-8-27b-medium	5.43% Reported settings Qwen3.8 27B (Xhigh); Model source: https://artificialanalysis.ai/models/qwen3-8-27b	Not reported	0.29% Reported settings Qwen3.8 27B (Non-reasoning); Model source: https://artificialanalysis.ai/models/qwen3-8-27b-non-reasoning	Not reported
EnterpriseOps-Gym-AAArtificial Analysis · Artificial Analysis published evaluation; see source for benchmark-specific settings Benchmark details Overall task success rate across all eight enterprise domains (oracle tool mode) · Benchmark developed by ServiceNow Research · Independently benchmarked by Artificial Analysis · Evaluation results measured independently by Artificial Analysis Unit: percent. Source collected 2026-10-18.	Not reported	Not reported	44.23% Reported settings Qwen3.8 27B (Xhigh); Model source: https://artificialanalysis.ai/models/qwen3-8-27b	Not reported	Not reported	Not reported
GDP.pdf: All-pass RateArtificial Analysis · Artificial Analysis published evaluation; see source for benchmark-specific settings Benchmark details Share of attempts where every atomic criterion passed · Independently benchmarked by Artificial Analysis · Evaluation results measured independently by Artificial Analysis Unit: percent. Source collected 2026-10-18.	10.4% Reported settings Qwen3.8 27B (Low); Model source: https://artificialanalysis.ai/models/qwen3-8-27b-low	12.8% Reported settings Qwen3.8 27B (Medium); Model source: https://artificialanalysis.ai/models/qwen3-8-27b-medium	16.4% Reported settings Qwen3.8 27B (Xhigh); Model source: https://artificialanalysis.ai/models/qwen3-8-27b	Not reported	5.2% Reported settings Qwen3.8 27B (Non-reasoning); Model source: https://artificialanalysis.ai/models/qwen3-8-27b-non-reasoning	Not reported
GDPval-AA v2.1 LeaderboardArtificial Analysis · Artificial Analysis published evaluation; see source for benchmark-specific settings Benchmark details Elo rating for performance on real-world work tasks · Anchored to DeepSeek V4.1 Flash (max) at 1600 · Higher is better · Evaluation results measured independently by Artificial Analysis Unit: elo. Source collected 2026-10-18.	1,362.53 Elo Reported settings Qwen3.8 27B (Low); Model source: https://artificialanalysis.ai/models/qwen3-8-27b-low	1,382.16 Elo Reported settings Qwen3.8 27B (Medium); Model source: https://artificialanalysis.ai/models/qwen3-8-27b-medium	1,419.88 Elo Reported settings Qwen3.8 27B (Xhigh); Model source: https://artificialanalysis.ai/models/qwen3-8-27b	Not reported	1,072.36 Elo Reported settings Qwen3.8 27B (Non-reasoning); Model source: https://artificialanalysis.ai/models/qwen3-8-27b-non-reasoning	Not reported
GPQA DiamondArtificial Analysis · Artificial Analysis published evaluation; see source for benchmark-specific settings Benchmark details Independently benchmarked by Artificial Analysis · Evaluation results measured independently by Artificial Analysis Unit: percent. Source collected 2026-10-18.	84.55% Reported settings Qwen3.8 27B (Low); Model source: https://artificialanalysis.ai/models/qwen3-8-27b-low	84.55% Reported settings Qwen3.8 27B (Medium); Model source: https://artificialanalysis.ai/models/qwen3-8-27b-medium	90.51% Reported settings Qwen3.8 27B (Xhigh); Model source: https://artificialanalysis.ai/models/qwen3-8-27b	Not reported	81.82% Reported settings Qwen3.8 27B (Non-reasoning); Model source: https://artificialanalysis.ai/models/qwen3-8-27b-non-reasoning	Not reported
Humanity's Last ExamArtificial Analysis · Artificial Analysis published evaluation; see source for benchmark-specific settings Benchmark details Independently benchmarked by Artificial Analysis · Evaluation results measured independently by Artificial Analysis Unit: percent. Source collected 2026-10-18.	14.04% Reported settings Qwen3.8 27B (Low); Model source: https://artificialanalysis.ai/models/qwen3-8-27b-low	14.09% Reported settings Qwen3.8 27B (Medium); Model source: https://artificialanalysis.ai/models/qwen3-8-27b-medium	33.92% Reported settings Qwen3.8 27B (Xhigh); Model source: https://artificialanalysis.ai/models/qwen3-8-27b	Not reported	12.14% Reported settings Qwen3.8 27B (Non-reasoning); Model source: https://artificialanalysis.ai/models/qwen3-8-27b-non-reasoning	Not reported
MLCR-AAArtificial Analysis · Artificial Analysis published evaluation; see source for benchmark-specific settings Benchmark details Share of tasks judged accurate, complete, and concise · Expert + compound (hard) sets · Benchmark developed by Wisedocs · Independently benchmarked by Artificial Analysis · Evaluation results measured independently by Artificial Analysis Unit: percent. Source collected 2026-10-18.	Not reported	Not reported	21.67% Reported settings Qwen3.8 27B (Xhigh); Model source: https://artificialanalysis.ai/models/qwen3-8-27b	Not reported	Not reported	Not reported
MMMU-ProArtificial Analysis · Artificial Analysis published evaluation; see source for benchmark-specific settings Benchmark details Independently benchmarked by Artificial Analysis · Evaluation results measured independently by Artificial Analysis Unit: percent. Source collected 2026-10-18.	73.76% Reported settings Qwen3.8 27B (Low); Model source: https://artificialanalysis.ai/models/qwen3-8-27b-low	74.16% Reported settings Qwen3.8 27B (Medium); Model source: https://artificialanalysis.ai/models/qwen3-8-27b-medium	76.3% Reported settings Qwen3.8 27B (Xhigh); Model source: https://artificialanalysis.ai/models/qwen3-8-27b	Not reported	69.94% Reported settings Qwen3.8 27B (Non-reasoning); Model source: https://artificialanalysis.ai/models/qwen3-8-27b-non-reasoning	Not reported
AA-Omniscience IndexArtificial Analysis · Artificial Analysis published evaluation; see source for benchmark-specific settings Benchmark details AA-Omniscience Index (higher is better) measures knowledge reliability and hallucination. It rewards correct answers, penalizes hallucinations, and has no penalty for refusing to answer. Scores range from -100 to 100, where 0 means as many correct as incorrect answers, and negative scores mean more incorrect than correct. · Evaluation results measured independently by Artificial Analysis Unit: index. Source collected 2026-10-18.	-26.67 Reported settings Qwen3.8 27B (Low); Model source: https://artificialanalysis.ai/models/qwen3-8-27b-low	-36.15 Reported settings Qwen3.8 27B (Medium); Model source: https://artificialanalysis.ai/models/qwen3-8-27b-medium	-9.98 Reported settings Qwen3.8 27B (Xhigh); Model source: https://artificialanalysis.ai/models/qwen3-8-27b	Not reported	-7.95 Reported settings Qwen3.8 27B (Non-reasoning); Model source: https://artificialanalysis.ai/models/qwen3-8-27b-non-reasoning	Not reported
AA-Omniscience AccuracyArtificial Analysis · 6000 questions; 1 attempt; correct-answer share; GPT-5.6 Luna medium grader Benchmark details AA-Omniscience Accuracy (higher is better) measures the proportion of correctly answered questions out of all questions, regardless of whether the model chooses to answer · Evaluation results measured independently by Artificial Analysis Unit: percent. Source collected 2026-10-18.	17.13% Reported settings Qwen3.8 27B (Low); Model source: https://artificialanalysis.ai/models/qwen3-8-27b-low	18.35% Reported settings Qwen3.8 27B (Medium); Model source: https://artificialanalysis.ai/models/qwen3-8-27b-medium	15.58% Reported settings Qwen3.8 27B (Xhigh); Model source: https://artificialanalysis.ai/models/qwen3-8-27b	Not reported	8.58% Reported settings Qwen3.8 27B (Non-reasoning); Model source: https://artificialanalysis.ai/models/qwen3-8-27b-non-reasoning	Not reported
SciCodeArtificial Analysis · Artificial Analysis published evaluation; see source for benchmark-specific settings Benchmark details Benchmark curated by scientists across 16 disciplines. SciCode is under review following an independent audit of the dataset · Independently benchmarked by Artificial Analysis · Evaluation results measured independently by Artificial Analysis Unit: percent. Source collected 2026-10-18.	40.05% Reported settings Qwen3.8 27B (Low); Model source: https://artificialanalysis.ai/models/qwen3-8-27b-low	39% Reported settings Qwen3.8 27B (Medium); Model source: https://artificialanalysis.ai/models/qwen3-8-27b-medium	46.64% Reported settings Qwen3.8 27B (Xhigh); Model source: https://artificialanalysis.ai/models/qwen3-8-27b	Not reported	36.23% Reported settings Qwen3.8 27B (Non-reasoning); Model source: https://artificialanalysis.ai/models/qwen3-8-27b-non-reasoning	Not reported
𝜏³-BankingArtificial Analysis · Artificial Analysis published evaluation; see source for benchmark-specific settings Benchmark details Benchmark developed by Sierra Research · Independently benchmarked by Artificial Analysis · Evaluation results measured independently by Artificial Analysis Unit: percent. Source collected 2026-10-18.	32.16% Reported settings Qwen3.8 27B (Low); Model source: https://artificialanalysis.ai/models/qwen3-8-27b-low	47.42% Reported settings Qwen3.8 27B (Medium); Model source: https://artificialanalysis.ai/models/qwen3-8-27b-medium	48.04% Reported settings Qwen3.8 27B (Xhigh); Model source: https://artificialanalysis.ai/models/qwen3-8-27b	Not reported	20% Reported settings Qwen3.8 27B (Non-reasoning); Model source: https://artificialanalysis.ai/models/qwen3-8-27b-non-reasoning	Not reported
Terminal-Bench-Science 0.1Artificial Analysis · mini-swe-agent; Terminal-Bench-Science 0.1.0; 70 tasks; pass@1 averaged over 3 repeats Benchmark details Share of tasks passed (pass@1), averaged over three repeats · Independently benchmarked by Artificial Analysis · Evaluation results measured independently by Artificial Analysis Unit: percent. Source collected 2026-10-18.	0.48% Reported settings Qwen3.8 27B (Low); Model source: https://artificialanalysis.ai/models/qwen3-8-27b-low	0.95% Reported settings Qwen3.8 27B (Medium); Model source: https://artificialanalysis.ai/models/qwen3-8-27b-medium	0.48% Reported settings Qwen3.8 27B (Xhigh); Model source: https://artificialanalysis.ai/models/qwen3-8-27b	Not reported	0% Reported settings Qwen3.8 27B (Non-reasoning); Model source: https://artificialanalysis.ai/models/qwen3-8-27b-non-reasoning	Not reported
Terminal-Bench v2.1Artificial Analysis · Artificial Analysis published evaluation; see source for benchmark-specific settings Benchmark details Benchmark developed by the Laude Institute and open-source contributors · Independently benchmarked by Artificial Analysis · Evaluation results measured independently by Artificial Analysis Unit: percent. Source collected 2026-10-18.	67.42% Reported settings Qwen3.8 27B (Low); Model source: https://artificialanalysis.ai/models/qwen3-8-27b-low	65.17% Reported settings Qwen3.8 27B (Medium); Model source: https://artificialanalysis.ai/models/qwen3-8-27b-medium	79.78% Reported settings Qwen3.8 27B (Xhigh); Model source: https://artificialanalysis.ai/models/qwen3-8-27b	Not reported	49.06% Reported settings Qwen3.8 27B (Non-reasoning); Model source: https://artificialanalysis.ai/models/qwen3-8-27b-non-reasoning	Not reported
Terminal-Bench v4.0Artificial Analysis · mini-SWE-agent v2.4.6; 66 tasks; pass@1 averaged over 3 repeats; 500 steps; task timeout up to 8 hours Benchmark details Benchmark developed by the Laude Institute, Stanford, and open-source contributors · Independently benchmarked by Artificial Analysis · Evaluation results measured independently by Artificial Analysis Unit: percent. Source collected 2026-10-18.	2.53% Reported settings Qwen3.8 27B (Low); Model source: https://artificialanalysis.ai/models/qwen3-8-27b-low	5.05% Reported settings Qwen3.8 27B (Medium); Model source: https://artificialanalysis.ai/models/qwen3-8-27b-medium	5.56% Reported settings Qwen3.8 27B (Xhigh); Model source: https://artificialanalysis.ai/models/qwen3-8-27b	Not reported	0% Reported settings Qwen3.8 27B (Non-reasoning); Model source: https://artificialanalysis.ai/models/qwen3-8-27b-non-reasoning	Not reported
ARC-AGI-2ARC Prize · v2 Semi-Private Benchmark details Original public board. Configuration and task coverage may differ; inspect source. Unit: percent. Source collected 2026-10-18.	22.78% Reported settings Qwen3.8-27B (Low); Source model ID: alibaba-qwen3-8-27b-low; 120 tasks	13.19% Reported settings Qwen3.8-27B (Medium); Source model ID: alibaba-qwen3-8-27b-medium; 120 tasks	42.36% Reported settings Qwen3.8-27B (XHigh); Source model ID: alibaba-qwen3-8-27b-xhigh; 120 tasks	1.53% Reported settings Qwen3.8-27B (None); Source model ID: alibaba-qwen3-8-27b-none; 120 tasks	Not reported	Not reported
VulcanBench v3 · Report 17VulcanBench · VulcanBench bare-bones API loop, deterministic hidden tests in Docker Benchmark details Frozen v3, 23 post-cutoff repository tasks. Published pass@1; repeat counts and time budgets vary by report. Display only pending comparability review. Unit: percent. Source collected 2026-10-18.	82.6% Reported settings Qwen3.8-27B; source label: low; low effort; 23 tasks; evaluation 2026-08-21. See source for repeat counts and wall-clock budget.	82.6% Reported settings Qwen3.8-27B; source label: medium; medium effort; 23 tasks; evaluation 2026-08-21. See source for repeat counts and wall-clock budget.	73.9% Reported settings Qwen3.8-27B; source label: xhigh; xhigh effort; 23 tasks; evaluation 2026-08-21. See source for repeat counts and wall-clock budget.	Not reported	Not reported	Not reported
Artificial Analysis Cyber IndexArtificial Analysis · Artificial Analysis Stirrup; defensive code audit, patch and end-to-end vulnerability tasks Benchmark details v1.0. Equal mean of CWE-Bench-AA, DeepsecBench-AA and CyberGym-E2E-AA. Display only. This composite measures defensive source-code tasks. Successful capability scores depend on safety refusals, and refusal rates are reported separately. It does not enter weighted Overall. https://artificialanalysis.ai/methodology/cyber-index Unit: index. Source collected 2026-10-18.	Not reported	Not reported	12.72 Reported settings Artificial Analysis Stirrup; v1.0; Qwen3.8 27B (xhigh); source label Qwen3.8 27B (Xhigh); mitigation standard; public model; source field cyberIndex; index 0–100; CWE-Bench-AA pass@1 0.266666666666667; refusal 0.6134453781512605; DeepsecBench-AA median F2, 3 repeats, GPT-5.6 Sol High judge 0.115062761506276; refusal 0.09090909090909091; CyberGym-E2E-AA pass@1 0; refusal 1; https://artificialanalysis.ai/models/qwen3-8-27b; page sha256 344ca66f5aa3f3abd1901b12bd29db73950d90e553fcb3ae900dcf590d7a6db2; encrypted payload sha256 b5e6ae259181eaed471e288c4bbdbd1088438e0a81caf19ec0fa60f7092d48f7; Display only. This composite measures defensive source-code tasks. Successful capability scores depend on safety refusals, and refusal rates are reported separately. It does not enter weighted Overall.	Not reported	Not reported	Not reported
Bug Hunt BenchBug Hunt Bench · Native agentic CLI per row, blind judge against a withheld key. The CLI is named in each configuration. Benchmark details 105 planted bugs across two production repositories (45 and 60). Score is (fixed / 105) * 100, counting planted bugs fixed only. Primary rows are featured and not superseded. Display only pending comparability review. Many rows are single runs. Unit: percent. Source collected 2026-10-18.	Not reported	Not reported	14.29% Reported settings Qwen3.8-27B; Claude Code / Alibaba API; effort xhigh (first_party). Mean of 3 runs; fixed 15 of 105 planted bugs. partial 0.7; claimed_only 4.3; extras 2.3; evaluation 2026-09-15. Score is (fixed / 105) * 100. Extras, partials and claimed-only are not scored. Source caveat: Every figure on this row is a MEAN of three runs, carried to one decimal - 15.0 of 105, range 4 (13 to 17). This row SUPERSEDES the single-run row (15/105) published 2026-09-12, which landed exactly on this mean by coincidence of one draw, [source caveat truncated].	Not reported	Not reported	Not reported
ThinkingBox-BenchMicrosoft and Hugging Face · ThinkingBox sandbox. Shared prompts, MCP tools, simulated user, and executable checks over terminal backend state. Benchmark details Task-weighted pass@1 on 507 stateful business workflows, 20 trials each. The stored score is the published overall from the 3 October 2026 blog. Display only. Excluded from weighted Overall. Unit: percent. Source collected 2026-10-18.	Not reported	Not reported	Not reported	Not reported	Not reported	51.7% Reported settings Qwen3.8-27B; ThinkingBox sandbox; shared prompts, tools, simulated user, and executable terminal-state checks; pass@1; 20 trials on each of 507 tasks; published task-weighted overall 51.70. Retail 64.03 (98); Auto insurance 47.85 (100); Travel 53.41 (104); Neobank 47.88 (104); Consulting 45.69 (101). Blog post 2026-10-03. The post does not name a reasoning-effort dial. Display only. Excluded from weighted Overall.
LMArena Agent ArenaLMArena · Arena Agent Mode. The board does not name a per-model scaffold. Benchmark details Aggregate Agent Arena IPS net improvement, stored as the board's printed percentage points. Arena Agent Mode does not name a per-model scaffold. Display only. Excluded from weighted Overall. Unit: index. Source collected 2026-10-18.	Not reported	Not reported	Not reported	Not reported	Not reported	-1.66 Reported settings Qwen 3.8 27B; Arena Agent Mode; IPS net improvement -1.66 points ±0.60. rank 37; observations 6972838; sessions 57343; published 2026-10-02. The board does not name a per-model scaffold. Display only. Excluded from weighted Overall.

What these comparisons mean

Each source reports its own effort labels. Matching labels do not establish equal compute budgets or identical fallback behavior. Open Reported settings to inspect the exact model name, agent and task coverage. Multiple results at one level remain visible rather than selecting the best. Source-reported None or Non-reasoning labels are not treated as evidence that an API supports those settings.

These measurements use their own dated source collection. The leaderboard fits identified configurations jointly; it contains no pooled model ratings. Explicit thinking budgets are separate configurations. Read the methodology.

Download all effort benchmark results · 8,916 results across 259 models · Collected 2026-10-18

Qwen3.8-27B · Medium

Compare with

One configuration, seven capabilities

Agentic

Hard reasoning

Coding

Human pref

Knowledge

Multimodal

Long context

Capability score evidence

Benchmark results