Netlify AI Gateway adds three frontier models, cuts an OpenAI price 33%

Netlify’s AI features pricing docs (docs.netlify.com) showed three new frontier-tier models added to the AI Gateway/Agent Runners catalog between the 2026-08-26 and 2026-09-07 captures: claude-fable-5-1 (Anthropic, $10.00 input / $50.00 output per 1M tokens, with a cheaper $0.25 cache-read than the pre-existing claude-fable-5’s $1.00), gemini-3.8-flash (Gemini, $0.75 input / $3.75 output — identical to the existing gemini-3.6-flash and gemini-3.7-flash rates), and gpt-6-astra (OpenAI, $10.00 input / $50.00 output / $1.00 cache-read / $12.50 cache-write — Netlify’s most expensive listed OpenAI model).

More consequentially for existing buyers, an already-listed OpenAI model was repriced downward by double digits: gpt-5.6-sol fell from $5.00 to $4.00 input (-20%), $0.50 to $0.40 cache-read (-20%), $30.00 to $20.00 output (-33%), and $6.25 to $5.00 cache-write (-20%) per 1M tokens. Since Netlify meters AI inference by converting the underlying model’s USD cost to credits (180 credits per $1), this repricing is directly bill-relevant for any team routing to gpt-5.6-sol through Netlify’s AI Gateway or Agent Runners.

The OpenRouter-routed portion of the catalog continued its pattern of rapid churn, growing from roughly 167 to roughly 173 models, with further double-digit moves on already-listed models: xiaomi/mimo-v2.5 output roughly doubled from $0.28 to $0.65 per 1M tokens; qwen/qwen2.5-vl-72b-instruct input more than tripled from $0.25 to $0.80 and output rose from $0.75 to $1.00; and z-ai/glm-5.2 output fell from $3.15 to $1.56. All other sampled Anthropic, Gemini, and OpenAI rates, and all four Netlify plan prices (Free $0, Personal $9, Pro $20, Enterprise Custom) and the 180-credits-per-$1 conversion, were unchanged. Source: docs.netlify.com AI features pricing page (screenshot diff, 2026-08-26 vs 2026-09-07 captures).