Qwen Thinking Models Spike +53.8% on OpenRouter; Goose Overtakes Agent Zero

Prices: OpenRouter hiked input/output on five Qwen thinking variants by +53.8% — qwen3-30b-a3b-thinking, qwen-plus-2025-07-28:thinking, qwen3-next-80b-a3b-thinking, qwen3-vl-235b-a22b-thinking, qwen3-vl-30b-a3b-thinking, qwen3-vl-8b-thinking (evts 29147,29146,29141,29140,29138,29137). qwen3-coder-next input jumped +63.6% (evt 29131). Elsewhere: minimax-m2.7 -17.1% (evt 29326), glm-5 -11% (evt 29323), nemotron-3-ultra-550b-a55b -16.7% (evt 29093), qwen3.7-max -15.3% (evt 29095). gemma-4-26b-a4b-it diverged: +14% on models.dev (evt 29328) vs -50% on OpenRouter (evt 29112). claude-fable-5 output ticked up +2.8% (evt 29303). Minor cuts hit hyper/qwen3-next-80b-a3b-instruct (-3.6%, evt 29331), qwen3-coder-480b (-5.3%, evt 29330), gpt-oss-120b (-1.1%, evt 29329).

Usage: goose overtook Agent Zero in OpenRouter coding-tool tokens, 13,824,406,432 vs 12,962,097,688 (7d, +6.7% margin) (evt 29542). north-mini-code crossed 50B OpenRouter tokens/day at 50,980,773,818 (evt 30622).

Releases: gpt-4o-mini-search-preview, inflection-3-productivity, and inflection-3-pi were delisted from OpenRouter after three absent snapshots (evts 30615, 30618, 30617).

Benchmarks: anthropic/claude-fable-5:batch landed top-5 on aa-coding-index (76.5), aa-intelligence-index (59.9), and aa-agentic-index (52.8) (evts 30557, 30556, 30558).

Plans: Replit delisted Core ($18/yr) and Pro ($90/yr) pricing (evt 29539); Lovable's Free tier price was delisted (evt 29537); Codex renamed Pro tiers to Pro (5x)/Pro (20x) (evt 29531); Kilo Code renamed its Pass tiers (evt 29536).