Skip to content
TechPP

Claude Fable 5.2 / Mythos 5.2 Benchmark Comparison: Tops Again!

Anthropic has finally announced its latest flagship models, Fable 5.2 and Mythos 5.2. Both models again see a significant improvement over their predecessors and top the table again. In this post, here is a full benchmark comparison of Fable 5.2 and Mythos 5.2 with Fable 5.1, Opus 5.5, GPT-6 Astra and Gemini 4 Carbon, along with pricing, specs and real-world outputs.

Benchmarks: Fable 5.1 vs Fable 5.2

BenchmarkClaude Fable 5.1Claude Fable 5.2Change
FrontierCode 2.032.4%Xhigh effort49.1%Xhigh effort+16.7 points
FrontierSWE v256.3%69.9%+13.6 points
AutomationBench31.4%51.4%+20.0 points
Terminal-Bench-Science 0.152.6%74.6%+22.0 points
WANDR68.7%82.1%+13.4 points
OSWorld 2.183.5%89.3%+5.8 points
RoboDojo Sim27.6%64.9%+37.3 points
AA-Briefcase v1.1 rubric61.5%66.1%+4.6 points
AA-Briefcase v1.1 Elo16691785+116
Remote Labor Index17.9%41.3%High effort+23.4 points
Humanity's Last Exam65.6%74.2%+8.6 points
Chartography44.8%73.4%+28.6 points

Fable 5.2 vs other models

BenchmarkClaude Fable 5.2Claude Opus 5.5GPT-6 AstraGemini 4 Carbon
FrontierCode 2.049.1%Xhigh effort45.5%42.8%41.1%Medium effort
FrontierSWE v269.9%62.3%65.5%61.2%
AutomationBench51.4%42.5%41.4%53.9%
Terminal-Bench-Science 0.174.6%75.3% on Mythos 5.258.7%68.1%64.8%
WANDR82.1%72.3%--
OSWorld 2.189.3%87.2%79.5%81.5%
RoboDojo Sim64.9%39.4%29.8%-
AA-Briefcase v1.1 rubric66.1%61.2%52.0%61.8%
AA-Briefcase v1.1 Elo1785180015691611
Remote Labor Index41.3%High effort21.3%20.8%-
Humanity's Last Exam74.2%67.7%57.2%-
Chartography73.4%64.4%71.0%72.5%

Fable 5.2 leads on 10 of these 12 results. Gemini 4 Carbon wins AutomationBench, and Opus 5.5 keeps a small lead in the AA-Briefcase Elo rating. A dash means no score was published for that model.

Mythos 5.2: cybersecurity and biology

BenchmarkClaude Mythos 5.2Claude Mythos 5.1Claude Opus 5.5GPT-6 Cyber
CyScenarioBench76.3%61.7%67.6%Without safeguards61.0%
CVE Rediscovery89.6%78.3%81.3%Without safeguards-
De novo binder design85.1%79.1%82.6%Without safeguards-

Fable 5.2 runs with cybersecurity and biology safeguards, so these scores are for Mythos 5.2, the same model for verified organizations. On CyScenarioBench, each lab ran its own set of tasks.

Pricing: Fable 5.1 vs Fable 5.2

PriceFable 5.1 and Mythos 5.1Fable 5.2 and Mythos 5.2
Input$10$8
Output$50$40
Cache reads$0.25$0.20
Cache writes, 5 minutes$12.50$10
Cache writes, 1 hour$20$16
Batch API$5 and $25$4 and $20
Fast modeNot offered$16 and $80Up to 2.5x faster generation
Output speedBaselineUp to 30% faster

Fable 5.2 and Mythos 5.2 cost 20% less than the 5.1 models on every token price.

Specs: Fable 5.2 vs Mythos 5.2

SpecClaude Fable 5.2Claude Mythos 5.2
Underlying modelSameSame
Who can use itAll API customersVerified organizationsCyber and Life Sciences Verification Programs
Cyber and biologyStrong safeguardsAbout 55% fewer blocked requests than Fable 5.1Expanded capabilities
Context window1M tokens1M tokens
Max output128K tokens128K tokens
InputText and imagesText and images
OutputTextText
ThinkingAdaptiveAlways onAdaptiveAlways on
Default effortHighHigh
LatencyModerateModerate
Knowledge cutoffJune 2026June 2026
Training cutoffJune 2026June 2026
ReleasedNovember 9, 2026November 9, 2026
API model IDclaude-fable-5-2claude-mythos-5-2
PlatformsClaude API, AWS, Google Cloud, Microsoft FoundryClaude API, AWS, Google Cloud, Microsoft Foundry

Claude API pricing

ModelInput, per 1M tokensOutput, per 1M tokens
Claude Haiku 5.5From $0.10From $0.50
Claude Sonnet 5.5$2$10
Claude Opus 5.5$4$20
Claude Fable 5.2$8$40
Claude Mythos 5.2$8$40
Claude Fable 5.1$10$50

Real-world outputs