Agent Index composite benchmark ranks Claude Opus 5.5 first at 90.4 for agentic work
anitakirkovska · reddit · 2026-11-06
A Reddit user built the Agent Index, a composite benchmark combining seven published evaluations covering browsing, terminal use, coding, document understanding, professional deliverables, multi-step processes and reasoning, normalized and weighted toward browsing and multi-step workflows. 26 models qualify.
November 2026 top 5: Claude Opus 5.5 (90.4), GPT-6 Astra (87.6, $30/M tokens), Claude Sonnet 5.5 (85.8, only $6/M), Claude Fable 5.1 (85.7), GPT-6.1 Sol (83.4).
Price-performance highlights (blended input/output $/M tokens): Sonnet 5.5 nearly matches Astra at one-fifth the price; GPT-6.1 Terra scores 67.4 at $3/M, matching the half-priced Kimi K3.1 Pro; Claude Haiku 5.5 is the budget pick at 45.4 for $0.30/M. The author curates published evals rather than running them, and asks agent builders to sanity-check the ranking against real workflows.
Related event: Homemade Agent Index Ranks Claude Opus 5.5 on Top(2 posts)→
More from Models
- Palisade's Jeff Ladish: models better at everything but still not funny or persuasive — JeffLadish · 2026-11-06
- SGLang ships day-0 Kimi K3.1 Pro support: 31K+ tok/s per GPU on GB300 NVL72 — ying11231 · 2026-11-06
- Palisade's Jeff Ladish calls to revive METR time-horizon estimates for new models — JeffLadish · 2026-11-06
- Leaker DM claims Qwen4 is dropping, but even the timeline is a guessing game — bclavie · 2026-11-06
- All 22 DeepSeek V4.1 Flash endpoints statistically in line, most consistent model in the Index — ArtificialAnlys · 2026-11-06
- All 22 DeepSeek V4.1 Flash endpoints match reference — 3.9-point range, most consistent model in the Index — ArtificialAnlys · 2026-11-06