Codebuff Sinks Below Six Rivals on OpenRouter Coding Tokens; DeepSeek, Venice Prices Swing

Usage Codebuff's 7d coding-token count collapsed to 52,027,900,800 from 117,125,611,935 prior, and it was overtaken by six apps: GDevelop (60,884,325,949, evt 53097), Studs.gg (62,779,172,352, evt 53095), OpenHands (65,924,587,732, evt 53094), Ito (78,231,721,651, evt 53093), Zed Editor (88,887,005,108, evt 53092) and Command Code (96,685,202,071, evt 53091). Separately Studs.gg passed Portkey AI, 62,779,172,352 vs 47,616,287,038 (evt 53096).

Prices DeepSeek-V4-Flash-0731 input cut $0.09→$0.08/MTok across OpenRouter, models.dev and DeepInfra listings (evts 52843, 52848, 52856). DeepSeek-V4-Pro input rose $0.435→$0.6864/MTok on models.dev (evt 52742) and $0.435→$0.6512 on OpenRouter (evt 52271). Qwen3.5-397B-A17B input jumped $0.39→$0.5/MTok, output $2.34→$3.6, on both OpenRouter (evt 52295) and models.dev (evt 52723). Venice's openai-gpt-56-luna-pro spiked $0.2667→$1.25/MTok (+368.7%, evt 52502). Requesty's gpt-5.6-luna@eu fell 80% to $0.22/MTok input (evt 52862); gpt-5.6-terra@eu fell 20% to $2.2 (evt 52863). Qwen3-Coder-480B input cut 21.8% to $0.445 on Hyper (evt 52683); Llama-3.3-70B input cut to $0.598 (evt 52669); Minimax-M2.7 input rose to $0.48 (evt 52655); GPT-OSS-120B input rose to $0.188 (evt 52677); Kimi-K2.5 input rose to $0.558 (evt 52645); Qwen3-Coder-30B output rose to $0.28 on both OpenRouter (evt 52317) and models.dev (evt 52709).

Plans Zencoder (evt 53090), Replit (evt 53088) and Emergent (evt 53085) dropped listed yearly prices for multiple tiers. Codex added Pro 5x/Pro 20x, removed Pro (evt 53083). Codebuff restructured its $100/$200/$500 monthly plans (evt 53082).

Benchmarks inclusionai/ling-3.0-tiny:free scored 16 on aa-agentic-index (evt 54200).