OpenRouter Purges Batch Endpoints; Cursor Overtakes Portkey on Coding Tokens

Prices: A dense sweep of models.dev repricings across gateways. Biggest mover: kilo/deepseek/deepseek-v4-pro input jumped $0.435→$1.6/MTok (+267.8%, evt 37144). Notable cuts: merge-gateway/openai/gpt-5.6-luna input $0.75→$0.2 (-73.3%, evt 37289); openrouter/z-ai/glm-5.2 input $0.4186→$0.2842 (-32.1%, evt 37170); deepinfra GLM-5.2 input $0.93→$0.75 (-19.4%, evt 37156). Notable hikes: kilo/google/gemma-3n-e4b-it input $0.02→$0.06 (+200%, evt 37152); kilo/google/gemma-3-27b-it input $0.03→$0.08 (+166.7%, evt 37150); kilo/deepseek/deepseek-r1-0528 input $0.45→$0.7 (+55.6%, evt 37146); merge-gateway/openai/gpt-5.6-sol input $3.75→$5 (+33.3%, evt 37288). Smaller repricings hit gemma-4-26b-a4b (evt 37277, 37147), gemma-4-31b (evt 37148), qwen3.6-27b (evt 37287), gpt-oss-120b (evt 37153), deepseek-chat-v3-0324 (evt 37143), deepseek-chat (evt 37142), r1-distill-llama-70b (evt 37145), and gpt-5.6-terra (evt 37290).

Usage: Cursor overtook Portkey AI in OpenRouter coding-token volume, 90,720,352,815 vs 83,261,786,719 tokens over 7 days (+9% margin, evt 37503).

Releases/Deprecations: OpenRouter delisted a wave of :batch endpoints absent for three straight daily snapshots, last seen 2026-07-31: Gemini 2.5 Pro, Claude Opus 4.1/4.5/4.6/4.7/4.8, Claude Sonnet 4.5/5, Claude Fable 5, GPT-5/5.1/5.2/5.4/5.4-mini/5.4-nano/5.5-mini/5-nano, Gemini 3.1 Pro Preview, and GPT-5.1-chat (evts 38574–38617).