Claude Haiku 4.5 vs Claude Fable 5 vs Claude Opus 4.8

Best cell per row highlighted. Null means undisclosed — never counted as zero.

Anthropic
Claude Haiku 4.5GA
8.4
AI Panel
Anthropic
Claude Fable 5GA
9.5
AI Panel
Anthropic
Claude Opus 4.8GA
9.2
AI Panel

Identity & lifecycle

ProviderAnthropicAnthropicAnthropic
Family / tierClaude 4 · HaikuClaude 5 · FableClaude 4 · Opus
StatusGAGAGA
Released2025-10-142026-06-082026-05-27
Knowledge cutoff2025-02

Architecture & context

Context window200K1M1M
Max output tokens64K128K128K
Input modalitiestext, imagetext, image, filetext, image
Reasoning modeoptionalalwaysalways
Open weightsNoNoNo

Pricing (per Mtok)

Input$1$10$5
Output$5$50$25
Cached input$0.1$1$0.5
Batch input$0.5$5$2.5
Free tierYesNoNo

Speed

Output speed (tok/s)101.860
Time to first token (s)0.860.5
Latency tierfastslowmedium

Trust & deployment

Trains on inputsNoNoNo
LicenseProprietaryProprietaryProprietary
CloudsBedrock, Vertex AI, Azure AI FoundryBedrock, Vertex AI, Azure AI FoundryBedrock, Vertex AI, Azure AI Foundry
ComplianceSOC2, HIPAA, GDPR, ISO27001SOC2, HIPAA, GDPR, ISO27001SOC2, HIPAA, GDPR, ISO27001

AI Panel scoring

Unified score8.49.59.2
Decision Maker8.59.59
Domain Strategist8.59.59
Finance Lead1078.5
Domain Practitioner8.59.59.5
Power User899
Skeptic7.57.58
Value score9.86.58

Benchmarks

MMLU-Pro76
GPQA Diamond67.293.6
SWE-bench Verified73.39588.6
Terminal-Bench41.758874.6
LMArena Elo1378
LMArena Coding Elo1436
Artificial Analysis Index31

Scores link to their sources. Missing cells mean the vendor hasn't disclosed a result — honesty over padding.