Did OpenAI flip OpenRouter share vs Anthropic from 20/80 to 50/50?

Opening

Baker’s OpenRouter share flip is the load-bearing claim — and his own reply undercuts a pure “quality won” read. @GavinSBaker says that per OpenRouter, OpenAI has gone from 20% share to 50% share versus Anthropic since June (Anthropic 80%→50%). @TaliaGold answers that token share and revenue share are diverging quickly; Baker replies that OpenAI is “for sure” discounting. @firstadopter layers product/policy (Astra performance + Anthropic data-retention) on top. Deep reply tail unread (top-replies); charts in Baker’s post not independently re-pulled here.

Sides

FOR — OpenAI regained relative OpenRouter share vs Anthropic (product + policy + price)

• @GavinSBaker [named] [book] — OpenRouter: OpenAI 20%→50% vs Anthropic since June
• @firstadopter [named] — attributes gains to strong performance and Anthropic’s “onerous data retention policy”; quotes Databricks @pwendell that Astra outperforms Opus 5 / Sol 5.6 on complex tasks
• @ramez [named] — OpenAI on a roll technically / UX / vibes; no obvious network effects or stickiness — only moat is staying ahead

THIRD — share metric may be dollars vs tokens; OpenAI is discounting

• @TaliaGold [named] — “Token share and revenue share diverging quickly” (quote of Baker)
• @GavinSBaker [named] [book] — “for sure OpenAI is discounting”

Room: tier-1 PM with chart; named operator (Databricks CTO/eng voice via firstadopter quote) on Astra; disagreement is less “is the chart fake” than “what mix of price vs quality vs policy.”

Quotes

Extraordinary share gains for OpenAI vs. Anthropic over the last two months.

Per Openrouter, OpenAI has gone from 20% share to 50% share vs. Anthropic (meaning Anthropic has gone from 80% to 50%) since June. — Gavin Baker (@GavinSBaker)

Token share and revenue share diverging quickly — Talia Goldberg (@TaliaGold)

@TaliaGold That is an interesting point - for sure OpenAI is discounting. — Gavin Baker (@GavinSBaker)

OpenAI is gaining market share due to its strong performance and Anthropic's onerous data retention policy.

“Astra unambiguously out performs our previous highest-end models (Opus 5, Sol 5.6) on highly complex tasks” — tae kim (@firstadopter)

1. OpenAI has been on a roll, technically, in user experience (fewer refusals), and vibes.

2. There continue to be no obvious network effects or even stickiness in AI. The only moat is staying ahead. This is an ultra-competitive, plural market, to customers' benefit. — Ramez Naam (@ramez)

What would settle it

OpenRouter (or equivalent) published monthly token-share and spend-share series for OpenAI vs Anthropic through October–November 2026 Knowable — OpenRouter Insights / lab disclosures — through 2026-10-31

Whether Anthropic reverses relative share after a model or policy change (retention, pricing, refusals) Knowable — next 4–8 weeks of OpenRouter / API mix prints — through 2026-10-31

Databricks or other large enterprise rollouts disclosing sustained Astra vs Claude coding spend (beyond N≈3500 engineer rollout note) Knowable — follow-on operator posts / earnings Q&A — Q4 2026

Delta

Prior archive doc tracked OpenRouter spend vs tokens concentration across three U.S. labs (discussion). New here: a bilateral OpenAI–Anthropic relative share path since June plus an explicit discounting admission from Baker.