Automated Audit Trail
RSS FeedBenchmark Changelog
Logged automatically by daily ingest workers when any score shifts beyond sensitivity thresholds or when new models are promoted.
Aug 16, 2026
Catalog refresh: Claude Opus 5 / Fable 5 / Sonnet 5, GPT-5.6 Sol·Terra·Luna, Gemini 3.6/3.7 Flash, Grok 4.6 Aug 12 prices, GLM-5.3 and Qwen3.8 27B (Aug 14).
Organic/GEO layer: What is Elo guide, answer-first boxes, FAQ schema, AI crawlers allowed, admin ingest/alias/promote desk. DeepSeek V4-Pro-0813 alias added. No official flagship after Aug 14.
Catalog add: Claude Haiku 4.5 ($1/$5 official) and Llama 4 Scout (open-weight long-context) for cheap-alternative compares.
Same-class predecessors: Claude 3 Opus / Opus 4 / 3.5 Sonnet / Sonnet 4 / 3.5 Haiku, GPT-4 Turbo, Grok 3, DeepSeek V3, Llama 3.1 70B, Gemini 1.5 Pro.
Aug 14, 2026
qwen-3-8-27bAlibaba published Qwen3.8 27B open weights.
glm-5-3Z.ai launched GLM-5.3 (post-train of GLM-5.2). Weights delayed for a safety review.
Aug 13, 2026
gemini-3-7-flashGoogle shipped Gemini 3.7 Flash at $0.75/$3.75 intro pricing.
Aug 12, 2026
grok-4-6xAI Grok 4.6 post-training refresh. Seed Elo/price snapshots updated.