Compare published benchmarks

Use the full performance tables published with model releases. Compare the tests both models have, with category weights you choose.

These are publisher-reported configurations. A shared table does not establish equal inference budgets or an independent reproduction. Win share measures the fraction of weighted test outcomes won, not the size or statistical significance of an advantage.

First model

Second model

Published evaluation

Category weights

Agentic Hard reasoning Coding Human pref Knowledge Multimodal Long context

Weights are relative. Categories share the total; each benchmark family shares its category equally. Related versions and settings divide their family’s allocation.

Include different reported configurations, with their differences shown

GPT-5.6 Sol and Kimi K3

No unambiguous shared results under these settings. Inspect the available results below or choose another source.

Missing results never count as losses. Changing the source, selected models or weights can change the comparison. This is not a global model rank.

Benchmark	GPT-5.6 Sol	Kimi K3	Comparison
ARC-AGI · 1 Hard reasoning · Higher is better	96.5% Reported settings & source ARC Prize semi-private validation; verified Fable5.1 max effort; comparator figures as reported in summary. Claude Fable 5.1 and Claude Mythos 5.1 System Card · Table8.1.A;8.16 · reviewed 2026-09-12	No result	Missing result
ARC-AGI · 2 Hard reasoning · Higher is better	92.5% Reported settings & source ARC Prize semi-private validation; verified Fable5.1 max effort; comparator figures as reported in summary. Claude Fable 5.1 and Claude Mythos 5.1 System Card · Table8.1.A;8.16 · reviewed 2026-09-12	No result	Missing result
BioMysteryBench · Human Solvable Knowledge · Higher is better	86.1% Reported settings & source Anthropic life-sciences evaluation; bash/editor/packages except protein design no tools; protocols use bash/editor/search; Understanding revised network restriction. Internal evaluation; not necessarily publicly released; compare only same task/grader revision. Percent rank correlation is not accuracy. Claude Fable 5.1 and Claude Mythos 5.1 System Card · 8.19 · reviewed 2026-09-12	No result	Missing result
BioMysteryBench · Human Difficult Knowledge · Higher is better	28.8% Reported settings & source Anthropic life-sciences evaluation; bash/editor/packages except protein design no tools; protocols use bash/editor/search; Understanding revised network restriction. Internal evaluation; not necessarily publicly released; compare only same task/grader revision. Percent rank correlation is not accuracy. Claude Fable 5.1 and Claude Mythos 5.1 System Card · 8.19 · reviewed 2026-09-12	No result	Missing result
ProteinGym · Hard Knowledge · Higher is better	35.5 percent rank correlation Reported settings & source Anthropic life-sciences evaluation; bash/editor/packages except protein design no tools; protocols use bash/editor/search; Understanding revised network restriction. Internal evaluation; not necessarily publicly released; compare only same task/grader revision. Percent rank correlation is not accuracy. Claude Fable 5.1 and Claude Mythos 5.1 System Card · 8.19 · reviewed 2026-09-12	No result	Missing result
Organic Chemistry · 2 revised Knowledge · Higher is better	43.2% Reported settings & source Anthropic life-sciences evaluation; bash/editor/packages except protein design no tools; protocols use bash/editor/search; Understanding revised network restriction. Internal evaluation; not necessarily publicly released; compare only same task/grader revision. Percent rank correlation is not accuracy. Claude Fable 5.1 and Claude Mythos 5.1 System Card · 8.19 · reviewed 2026-09-12	No result	Missing result
Protocols · Troubleshooting Knowledge · Higher is better	56.4% Reported settings & source Anthropic life-sciences evaluation; bash/editor/packages except protein design no tools; protocols use bash/editor/search; Understanding revised network restriction. Internal evaluation; not necessarily publicly released; compare only same task/grader revision. Percent rank correlation is not accuracy. Claude Fable 5.1 and Claude Mythos 5.1 System Card · 8.19 · reviewed 2026-09-12	No result	Missing result
Protocols · Understanding network-restricted Knowledge · Higher is better	63.9% Reported settings & source Anthropic life-sciences evaluation; bash/editor/packages except protein design no tools; protocols use bash/editor/search; Understanding revised network restriction. Internal evaluation; not necessarily publicly released; compare only same task/grader revision. Percent rank correlation is not accuracy. Claude Fable 5.1 and Claude Mythos 5.1 System Card · 8.19 · reviewed 2026-09-12	No result	Missing result
FrontierSWE · 2 Coding · Higher is better	0.32 fraction Reported settings & source Proximal agent harness; max effort; 34 tasks, five trials/task; mean score on 0..1 scale. Claude Fable 5.1 and Claude Mythos 5.1 System Card · 8.5 · reviewed 2026-09-12	No result	Missing result
Terminal-Bench · 4.0 Agentic · Higher is better	37.3% Reported settings & source Claude Code --bare max effort, 15 trials/task over66tasks for Claude; GPT Codex CLI max from public board. Uses more precise section values rather than rounded headline table. Distinct from2.1; timeouts/resources changed. Comparison limit: System card section 8.6 cites the public Codex CLI result for Sol, while the Claude rows are internal Claude Code --bare reruns. Different agent harnesses and runs cannot form a matched comparison. Claude Fable 5.1 and Claude Mythos 5.1 System Card · 8.6 · reviewed 2026-09-12	No result	Missing result
Terminal-Bench-Science · 0.1 Agentic · Higher is better	22.4% Reported settings & source 70tasks; Claude Code --bare max; Fable10trials/task, Opus12; GPT Codex CLI max from public board. Claude Fable 5.1 and Claude Mythos 5.1 System Card · 8.7 · reviewed 2026-09-12	No result	Missing result
SWE-bench Pro Coding · Higher is better	64.6% Reported settings & source Anthropic reported configuration; adaptive thinking at max effort, default sampling, five-trial mean unless section specifies otherwise; context at most 1M. Comparator configuration follows cited prior card or board. Agent identity is not established as mini-swe-agent. Claude Fable 5.1 and Claude Mythos 5.1 System Card · Table 8.1.A; section 8.2 · reviewed 2026-09-12	No result	Missing result
CursorBench · 3.2.0 Coding · Higher is better	67.2% Reported settings & source Cursor production agent harness; independently measured by Cursor; max effort. Claude Fable 5.1 and Claude Mythos 5.1 System Card · 8.8 · reviewed 2026-09-12	No result	Missing result
GDPval-AA · 2 Human pref · Higher is better	1711 Reported settings & source Artificial Analysis independent agentic shell/web evaluation;220GDPvalgoldtasks; blind pairwise Elo; max effort for Claude. Claude Fable 5.1 and Claude Mythos 5.1 System Card · Table8.1.A;8.15.3 · reviewed 2026-09-12	No result	Missing result
AA-Briefcase Human pref · Higher is better	1502 Reported settings & source Artificial Analysis long-horizon knowledge projects; rubric and panel pairwise judging; Claude max effort. Claude Fable 5.1 and Claude Mythos 5.1 System Card · Table8.1.A;8.15.4 · reviewed 2026-09-12	No result	Missing result
AutomationBench Agentic · Higher is better	19.6% Reported settings & source Private held-out board; simulated business-workflow app APIs; all assertions must pass; max effort stated for Fable5.1/Opus5. Claude Fable 5.1 and Claude Mythos 5.1 System Card · Table8.1.A;8.15.6 · reviewed 2026-09-12	No result	Missing result

Explore all benchmark coverage · Compare the calibrated benchmark set · Download all observations