VerdictPal
Join

Models / Anthropic

Claude Fable 5

Claude Fable · Released 2026-06-09

Anthropic's June 2026 Mythos-class GA release. #1 on the Vals Index (75.15%) and Vals Multimodal Index (74.15%). 1M context, 128K output, safety classifiers route high-risk queries to Opus 4.8. $10 / $50 per 1M tokens.

Long-horizon agentic work — multi-day coding migrations, finance analysis, vision-heavy rebuilds, and research pipelines that outlast Sonnet or Opus sessions.

#1 of 21 on Vals Index · current snapshot
75 Vals Index · 14 benchmark rows · 4 sourcesPosition is benchmark-specific — not a cross-family or cross-source ranking.
Composite72
Context1M
Input / 1M$10
Output / 1M$50
Knowledge cutoff2026-04-01
Output speed70 t/s
TTFT (default API)68.33s
Statusflagship
ClaudeFactory
ReasoningCodingAgenticLong context
Reasoning
93
Coding
70
Agentic
100
Long context
70

83average across 4 tested families

Editor's note
Fable 5 is the model we reach for when a task needs days of autonomous work, not minutes. The safety classifiers are conservative — cyber, biology, and chemistry prompts may fall back to Opus 4.8 — so read the system card before pinning it on sensitive workflows. The Vals 75.15% is the highest public intelligence row we track.
Benchmark placements

Where Claude Fable 5 places on each public benchmark source that publishes a row. Rank counts every model with a latest row in the same test — not a universal quality score.

BenchmarkFamilyRankScoreSourceDate
Artificial Analysis Intelligence IndexReasoning#1/ 6760Artificial Analysis2026-07-21
DeepSWECoding#1/ 670DeepSWE leaderboard v1.12026-06-20
GDPvalAgentic#1/ 6100Artificial Analysis — GDPval-AA v22026-06-17
Humanity's Last ExamReasoning#1/ 6553Artificial Analysis2026-07-21
SciCodeCoding#1/ 6560Artificial Analysis2026-07-21
Vals IndexAgentic#1/ 2075Vals AI — Vals Index2026-07-16
Terminal-BenchAgentic#4/ 6985Artificial Analysis2026-07-21
Vending-Bench 2Agentic#5/ 5Andon Labs Vending-Bench 22026-06-01
AA time to first tokenPerformance#7/ 69Artificial Analysis2026-07-21
GPQA DiamondReasoning#8/ 6993Artificial Analysis2026-07-21
τ³-Bench BankingAgentic#15/ 6462Artificial Analysis — τ³-Bench Banking2026-06-17
AA-LCRLong context#22/ 6470Artificial Analysis2026-07-21
IFBenchReasoning#39/ 5763Artificial Analysis2026-07-21
AA output speedPerformance#40/ 69Artificial Analysis2026-07-21
Family context

The three highest-scoring carded models in each capability family. Where Claude Fable 5 shows up, it's highlighted.

ReasoningClaude Fable 5 · 93
  1. 1
    Claude Mythos Preview
    95
  2. 2
    GPT-5.6 Sol
    94
  3. 3
    Kimi K3
    94
CodingClaude Fable 5 · 70
  1. 1
    GPT 5.5
    96
  2. 2
    Claude Opus 4.8
    96
  3. 3
    Claude Opus 4.6
    96
AgenticClaude Fable 5 · 100
  1. 1
    GPT 5.5
    100
  2. 2
    Claude Opus 4.8
    100
  3. 3
    Claude Fable 5
    100
Long contextClaude Fable 5 · 70
  1. 1
    Kimi K3
    75
  2. 2
    GPT-5.1
    75
  3. 3
    GPT-5.6 Sol
    74
Newest receipts

Benchmark rows added to the public ledger for Claude Fable 5 in the last 120 days. Older rows live in the full table below.

14 recent
BenchmarkFamilyScoreSourceDays ago
AA-LCRLong context70%Artificial Analysis2
AA output speedPerformance70 t/sArtificial Analysis2
AA time to first tokenPerformance68.33sArtificial Analysis2
Artificial Analysis Intelligence IndexReasoning59.9%Artificial Analysis2
GPQA DiamondReasoning92.6%Artificial Analysis2
Humanity's Last ExamReasoning53.3%Artificial Analysis2
IFBenchReasoning63.47%Artificial Analysis2
SciCodeCoding60.2%Artificial Analysis2
Terminal-BenchAgentic84.64%Artificial Analysis2
Vals IndexAgentic75.1%Vals AI — Vals Index7
DeepSWECoding70%DeepSWE leaderboard v1.133
GDPvalAgentic1818Artificial Analysis — GDPval-AA v236
τ³-Bench BankingAgentic62.4%Artificial Analysis — τ³-Bench Banking36
Vending-Bench 2Agentic5680.26Andon Labs Vending-Bench 252
Family coverage

A high composite that hides a weak family is a trap. These bars surface the families where this model hasn't been publicly tested, and where it leads.

Reasoning93
Coding70
Agentic100
Long context70
Price vs. performance
0255075100$0.10$0.30$1$3$10$30input USD / 1M tokens · log scale →← intelligence

Hover or tab any dot for name, score, and input price.

AnthropicOpenAIMistralAlibabaDeepSeekGoogleMoonshotMiniMaxxAIMetaZhipuXiaomiMeituanCohere

Full benchmark ledger

Every catalog benchmark for Claude Fable 5. Scores link to the original source; gaps mean no public row exists yet.

14 sourced rows

14 of 32 catalog benchmarks have a sourced row for Claude Fable 5.

Reasoning

Coding

Agentic

Long context

Performance

Changelog
  1. Released Claude Fable 5 — first GA Mythos-class model; Vals Index 75.15%.
Sources