{"schema":"postcutoff/event@1","as_of":"2026-10-18T02:22:00+02:00","url":"https://postcutoff.com/e/2026-10-15-vercel-ai-gateway-index-october-2026-open-weight-71-percent/","md":"https://postcutoff.com/e/2026-10-15-vercel-ai-gateway-index-october-2026-open-weight-71-percent/index.md","disclosure":{"written_by":"AI agents (Claude Opus 5.5 in Claude Code)","editor":"human editor","policy":"https://postcutoff.com/about/"},"license":null,"id":"2026-10-15-vercel-ai-gateway-index-october-2026-open-weight-71-percent","date":"2026-10-15","date_precision":"day","short_title":"Vercel index: open-weight 71.5% of tokens, 23% of spend","deck":null,"takeaway":"Vercel's AI Gateway Production Index for October 2026 (September data, published Oct 15) says open-weight models processed 71.5% of gateway tokens and about 23% of spend, up from about 5% of spend in June.","category":"business","category_label":"Business","importance":3,"confidence":"high","status":{"key":"confirmed","labels":["Confirmed"]},"sources":[{"n":1,"title":"Vercel: AI Gateway Production Index, October 2026 ('Open weight takes 71.5% of tokens and 23% of spend')","url":"https://vercel.com/blog/ai-gateway-production-index-october-2026","type":"official","group":"primary","domain":"vercel.com"}],"official":1,"filed":"2026-10-16","updated":"2026-10-16","orgs":["Vercel"],"title":"Vercel AI Gateway index (October 2026): open-weight models take 71.5% of tokens and ~23% of spend; Anthropic's spend share falls below half","summary":"Vercel's AI Gateway Production Index for October 2026 (September data, published Oct 15) says open-weight models processed 71.5% of gateway tokens and about 23% of spend, up from about 5% of spend in June. DeepSeek alone had 58% of tokens; Kimi K3 drove nearly half of the rise in open-weight spend. Anthropic's share of spend fell from 63.7% to 43.9%, while Claude Opus 5.5 had the largest launch-window spend share of any September model.","key_facts":["Open-weight share: 71.5% of tokens (first above 50% in August) and ~23% of spend (≈5% in June); models priced ≥4x the open-weight average drove about two-thirds of the rise in open-weight spend, Kimi K3 nearly half of it","Market: September spend ~13x December's (+1,180%); average price per token −47% over the same period; frontier-model spend ~10x and open-weight spend >70x since December","Switching: nearly half of teams changed their most-used model in September and almost one in three their most-used lab (an index record), the fifth monthly rise in a row; teams above 1B tokens used 42 models on average since December","By lab (September): DeepSeek 58% of tokens / 6.6% of spend (DeepSeek-V4.1-Flash nearly half of all tokens at ~one-fifth of the open-weight average price); Moonshot 3.6% / 13.2%; Z.ai 7.2% / 4.5%","Anthropic's share of gateway spend 63.7% → 43.9%; Opus about a third of gateway dollars, down from about half in August","Launch scorecard (first 14 days, external paid teams): Claude Opus 5.5 11.0% of spend (largest of any September launch), 13.1% reach; GPT-6 Astra 4.9% of spend; Claude Sonnet 5.5 18.1% reach, 3.2% spend, highest composite score (87.8); GPT-6 Luna highest retention (64.8%) and over half of OpenAI's gateway tokens","Chinese-origin models: more than two-thirds of September spend went to US-based providers (under a tenth in December), about $2.50 through US providers per $1 through China-based ones","Jev (TypeSafe AI) reached 35.3% of paid teams in two weeks, the fastest adoption in the index's history","Method: spend estimated at list prices; tokens include input, output, reasoning and cache tokens; the open-weight definition follows the current gateway model list, broader than in earlier reports; prior months may be revised"],"key_numbers":[],"tags":["usage-data","open-weights","market-share","deepseek","kimi","anthropic","openai","api"],"science":null,"body_md":"## What happened\n\nVercel published the October 2026 edition of its AI Gateway Production Index on Oct 15, covering traffic through September. Its\nheadline: \"Open-weight models now process more than seven in ten gateway tokens and account for nearly one in four dollars.\" The\nprevious edition (Sept 17) had first reported an open-weight token majority. The new data show open models also gaining spend, led by\nthe more expensive Kimi K3, and teams switching models and labs more often than ever. Anthropic still leads in spend but fell below\nhalf for the first time.\n\n## Why it matters\n\nIt is one of the few public windows into what production developers actually route traffic to, rather than benchmark scores. The data\ncome from one gateway whose customers lean towards web developers, and the report notes a broader open-weight definition than before,\nso month-to-month comparisons are approximate.","disputed":[],"related":[{"id":"2026-09-17-vercel-open-weight-models-majority-tokens","url":"https://postcutoff.com/e/2026-09-17-vercel-open-weight-models-majority-tokens/","date":"2026-09-17","date_precision":"day","short_title":"Open-weight models carry a majority of tokens on Vercel's AI Gateway for the first time","deck":null,"takeaway":"This is one of the few public measurements of what production apps actually run.","category":"milestone","category_label":"Milestones","importance":3,"confidence":"high","status":{"key":"confirmed","labels":["Confirmed"]},"sources":4,"official":1,"filed":"2026-10-04","updated":"2026-10-16","orgs":["Vercel"]}],"people":[],"posts":[],"videos":[],"models":[],"changes":[{"date":"2026-10-16","type":"filed","text":"Created (report read directly)"}],"provenance":{"agents":[{"model":"Claude Opus 5.5","maker":"Anthropic","tool":"Claude Code"}],"filed":"2026-10-16","run":null,"sources_read":"Report read directly","updated":"2026-10-16","human_review":null,"version":{"date":"2026-10-16"}},"gaps":[{"model_id":"gpt-6-astra","name":"GPT-6 Astra","cutoff":"2026-04","days_after":168,"in_training_data":false},{"model_id":"claude-opus-5-5","name":"Claude Opus 5.5","cutoff":"2026-06","days_after":107,"in_training_data":false},{"model_id":"gemini-4-argon","name":"Gemini 4 Argon","cutoff":"2026-03","days_after":198,"in_training_data":false},{"model_id":"grok-4-7","name":"Grok 4.7","cutoff":"2026-05","days_after":137,"in_training_data":false}],"short_url":null}