Expert-authored questions stretch academic reasoning and knowledge across many disciplines.
99 model results
One row per model and compatible score measure. Mean = equal-weight arithmetic average of imported reports. Distinct configurations reported for the same model in a source table remain separate. Reports can overlap; the range is not a confidence interval. Unbundle to inspect each observation.
61.35% · Humanity’s Last Exam · AA implementation [1]
Published observations span 6.35 percentage points. Configuration and source differences contribute to this spread; it is not measured run-to-run uncertainty.
Published observations span 7.3 percentage points. Configuration and source differences contribute to this spread; it is not measured run-to-run uncertainty.
#3 of 87 in this displayed comparison; 3.36 pp below the leader.
Published observations span 19.1 percentage points. Configuration and source differences contribute to this spread; it is not measured run-to-run uncertainty.
#4 of 87 in this displayed comparison; 5.47 pp below the leader.
54.68% · Humanity’s Last Exam · AA implementation [1]
Published observations span 2.52 percentage points. Configuration and source differences contribute to this spread; it is not measured run-to-run uncertainty.
#6 of 87 in this displayed comparison; 8.56 pp below the leader.
Published observations span 8.84 percentage points. Configuration and source differences contribute to this spread; it is not measured run-to-run uncertainty.
#10 of 87 in this displayed comparison; 13.99 pp below the leader.
Published observations span 7 percentage points. Configuration and source differences contribute to this spread; it is not measured run-to-run uncertainty.
#14 of 87 in this displayed comparison; 16.63 pp below the leader.
Published observations span 13 percentage points. Configuration and source differences contribute to this spread; it is not measured run-to-run uncertainty.
#16 of 87 in this displayed comparison; 18.03 pp below the leader.
Published observations span 19.3 percentage points. Configuration and source differences contribute to this spread; it is not measured run-to-run uncertainty.
#22 of 87 in this displayed comparison; 20.18 pp below the leader.
Published observations span 0.73 percentage points. Configuration and source differences contribute to this spread; it is not measured run-to-run uncertainty.
#28 of 87 in this displayed comparison; 21.56 pp below the leader.
Published observations span 15.86 percentage points. Configuration and source differences contribute to this spread; it is not measured run-to-run uncertainty.
#29 of 87 in this displayed comparison; 21.81 pp below the leader.
Published observations span 0.26 percentage points. Configuration and source differences contribute to this spread; it is not measured run-to-run uncertainty.
#30 of 87 in this displayed comparison; 22.4 pp below the leader.
Published observations span 8.3 percentage points. Configuration and source differences contribute to this spread; it is not measured run-to-run uncertainty.
#31 of 87 in this displayed comparison; 22.88 pp below the leader.
Published observations span 19.6 percentage points. Configuration and source differences contribute to this spread; it is not measured run-to-run uncertainty.
#33 of 87 in this displayed comparison; 24.53 pp below the leader.
Published observations span 19.6 percentage points. Configuration and source differences contribute to this spread; it is not measured run-to-run uncertainty.
#36 of 87 in this displayed comparison; 26.03 pp below the leader.
Published observations span 12.6 percentage points. Configuration and source differences contribute to this spread; it is not measured run-to-run uncertainty.
#38 of 87 in this displayed comparison; 27.43 pp below the leader.
Published observations span 25.83 percentage points. Configuration and source differences contribute to this spread; it is not measured run-to-run uncertainty.
#39 of 87 in this displayed comparison; 27.53 pp below the leader.
Published observations span 17.7 percentage points. Configuration and source differences contribute to this spread; it is not measured run-to-run uncertainty.
#41 of 87 in this displayed comparison; 28.26 pp below the leader.
Humanity’s Last Exam · AA implementation · 2026-10-26
Publisher-selected recent-model cohort from the public evaluation page. Model labels preserve effort and fallback configurations. Values are a dated snapshot, not a live feed. Consult the publisher’s methodology for the harness, task split and judging; do not merge this cohort with vendor-run results.
Values retain the model card’s original formatting and score units. HLE is without tools; the final three visual evaluations use tools. The dagger marks source qualifications: inspect the original report before treating those results as matched.
Reported values are percentages except GDPval-AA, which is Elo. Tool-enabled and partial scores keep their labels. Effort settings, fallback models, and source-run protocols vary; see the original footnotes.
Launch-reported maxima across reasoning efforts, not a uniformly matched effort comparison. Research/API settings can differ from ChatGPT. Source footnotes disclose altered harnesses, model revisions and reproduced competitor results. A dagger marks cells with additional source qualifications. Preserve each row’s named subset and tools.
Earlier-2026 baseline cohort from the official model card. Asterisks identify competitor scores re-evaluated by Kimi rather than cited from their reports. Coding scores average ten runs; vision averages three runs. Tool-enabled and no-tools results, F1 and exact accuracy, pass@3 and pass^3 remain separate rows. MCPMark here predates the Verified default. APEX uses 452 public tasks.
HLE-Full · 2026-10-26
Earlier-2026 baseline cohort from the official model card. Asterisks identify competitor scores re-evaluated by Kimi rather than cited from their reports. Coding scores average ten runs; vision averages three runs. Tool-enabled and no-tools results, F1 and exact accuracy, pass@3 and pass^3 remain separate rows. MCPMark here predates the Verified default. APEX uses 452 public tasks.
Earlier-2026 baseline comparison from the official Qwen model card. Original model labels and score formatting are retained. Language and vision tables use different Qwen reference models. Tools, image inputs, agent scaffolds and task subsets can differ; follow the model card’s evaluation details. Video-MME subtitle settings and HLE tool settings stay in separate rows. Original MCP-Mark results predate MCPMark Verified.
HLE w/ tool · Qwen report · 2026-10-26
Earlier-2026 baseline comparison from the official Qwen model card. Original model labels and score formatting are retained. Language and vision tables use different Qwen reference models. Tools, image inputs, agent scaffolds and task subsets can differ; follow the model card’s evaluation details. Video-MME subtitle settings and HLE tool settings stay in separate rows. Original MCP-Mark results predate MCPMark Verified.
Epoch AI public data snapshot, CC BY 4.0. Externally collected result; original evaluation provider is linked. CSV: hle_external.csv; score column: Accuracy; fraction scale: 1. Model version is retained in configuration. Original source authors retain credit and any additional licensing terms.
Humanity’s Last Exam · HLE-Full w/ tools · Kimi K3.1 Pro · complete release comparison · 2026-10-27
Published model comparison, 2026. Kimi K3.1 Pro · complete release comparison. Exact values from the publisher’s embedded full comparison table. Maximum effort, fallback settings and dated model versions remain in model labels. Blank cells stay missing. Source footnotes specify mixed harnesses and externally quoted comparisons; this is a launch-reported cohort rather than a new independent evaluation. Benchmarks · full embedded comparison table · row 11; source precision and any reported uncertainty retained. These are source-reported observations, not evaluations run by RLTasks.
Humanity’s Last Exam · HLE-Full · Kimi K3.1 Pro · complete release comparison · 2026-10-27
Published model comparison, 2026. Kimi K3.1 Pro · complete release comparison. Exact values from the publisher’s embedded full comparison table. Maximum effort, fallback settings and dated model versions remain in model labels. Blank cells stay missing. Source footnotes specify mixed harnesses and externally quoted comparisons; this is a launch-reported cohort rather than a new independent evaluation. Benchmarks · full embedded comparison table · row 37; source precision and any reported uncertainty retained. These are source-reported observations, not evaluations run by RLTasks.
TRACE THE SCORE · Per-model notes use published breakdowns where available.
Score vs. release date
Epoch AI · 56 observations
HLE · Accuracy · % · higher is better
Dates are model release dates recorded by Epoch AI, not evaluation dates. Each point retains its published score and configuration. Separate releases are not a matched longitudinal experiment. Missing dates are omitted. Explore all reported results ↓
The dashed line traces the best observed score through the model release dates. It describes this selection of reports, with their original configurations. It does not estimate unmeasured models or costs.
View plotted values and sources (56)
Humanity’s Last Exam · HLE · Accuracy · %. Published measurements in their original score units.
These model reports include this benchmark. Their harnesses, task subsets, scoring policies, and resource budgets can differ; follow each report’s evaluation notes before joining the numbers.
Expert-authored questions stretch academic reasoning and knowledge across many disciplines.
How to read the result
Use the subject breakdown to see whether gains are broad or concentrated in a narrow specialty.
Limits and comparison traps
Text-only subsets, full multimodal inputs, tool access, and answer grading define different evaluation settings.
Epoch data and evaluation setting
Epoch AI publishes this benchmark in its capability collection. Imported rows retain the published score column and model configuration. Original authors and evaluation providers remain credited in the result references.
Epoch attribution and score coverage
Data is attributed to Epoch AI under CC BY 4.0, with original evaluation providers credited in the linked source snapshots. Scores are imported only where a published mapping identifies the field and its scale; missing or incompatible records are omitted rather than assigned zero.
Published evaluation
Exact values from the publisher’s embedded full comparison table. Maximum effort, fallback settings and dated model versions remain in model labels. Blank cells stay missing. Source footnotes specify mixed harnesses and externally quoted comparisons; this is a launch-reported cohort rather than a new independent evaluation.
Where this fits in an evaluation suite
Ask what the task actually requires: recall, deduction, search, or a mixture. Separate closed-book reasoning from retrieval and tool-assisted problem solving.
Start with a capability hypothesis, then choose a second test with a different failure mode. Keep the checkpoint, reasoning effort, agent tools, sampling policy, and scoring procedure fixed when measuring a training intervention. Record failures and missing outputs in the denominator.
A reproducible comparison
Exact values from the publisher’s embedded full comparison table. Maximum effort, fallback settings and dated model versions remain in model labels. Blank cells stay missing. Source footnotes specify mixed harnesses and externally quoted comparisons; this is a launch-reported cohort rather than a new independent evaluation.
Publish the exact run configuration, item coverage, score units, and available uncertainty. If a source supplies only an aggregate or a single run, preserve that limit. A new model name or a rounded score alone does not establish a capability improvement.