Comparação de modelos

Escolha até 4 modelos e compare-os em todas as métricas que acompanhamos.

MétricaClaude Mythos PreviewClaude Fable 5Claude Fable 5.1
ProvedorAnthropicAnthropicAnthropic
LicençaProprietaryProprietaryProprietary
Podium Score97.493.488.4
Posição geral12999
Elo de arena1509
Índice de Inteligência60.0
coding Coding80.984.195.5
math Math100.0
reasoning Reasoning98.881.092.0
agentic Agentic96.2
knowledge Knowledge100.043.358.0
multimodal Multimodal100.062.0
long-context Long Context99.2
speed Speed10.245.0
value Value71.475.0
Velocidade80 t/s71 t/s85 t/s
Tempo até o primeiro token141 ms350 ms
Preço de entrada ($/M)$3/M$10/M$3/M
Preço de saída ($/M)$15/M$50/M$15/M
Janela de contexto1000K1000K500K
SWE-Bench Verified93.9%95%
GPQA Diamond94.6%92.6%
Multilingual MMLU92.7%
Browsecomp86.9%
CharXiv Reasoning93.2%
Humanity’s Last Exam64.7%53.3%
SWE-Bench Pro77.8%
SciCode60.2%
LiveCodeBench70%
Ifbench63.5%
τ²-Bench Retail98.5%
Terminal-Bench Hard62.9%
FrontierMath
OSWorld
MMMU-Pro
MRCR v2
MCP Atlas
Toolathlon
Apex Agents
ScreenSpot-Pro
AIME 2025
ARC-AGI-2
Mmlupro
SimpleQA
MMLU-Pro
Ifeval
MATH
Aime2025
Math500
GpqaDiamond
Mmlu
IntelligencePerDollar
OutputSpeed
ContextWindowSize
ArenaHard
MMMU
HumanEval
0 / 4