model routing

Google has removed manual model selection for free Gemini app users, leaving 'Auto' as the only option. Most prompts will go to Flash-Lite, with harder ones routed to Flash or Pro, according to Google. New thinking levels (Low, Medium, High) are also rolling out to subscribers.

TypeSafe, founded by ChatGPT co-inventor Diogo Almeida, launched Jev on September 15, 2026 — a model that cannot generate free-form text but claims to classify, route, and score 20-200x faster and 40-400x cheaper than small frontier LLMs. Trained via a new method called RLCD, Jev targets production systems that use LLMs purely as structured judges or routers.

OpenRouter has introduced DeepSeek: DeepSeek Pro Latest, a routing alias that automatically points to whichever DeepSeek Pro model is newest. The endpoint offers a 1,049K token context window at $0.58 per 1M input tokens and $1.74 per 1M output tokens.

Sakana AI has released Fugu Ultra v1.1, an update to its multi-model router, claiming performance gains up to 7.9 points over v1.0 and results that beat Anthropic's Fable 5 despite Fable 5 not being part of the router's model pool. All benchmark figures come from Sakana itself and remain unverified.

OpenRouter has released Fusion, a multi-model routing system that processes prompts through parallel expert model panels with web search enabled, then uses a judge model to synthesize consensus, contradictions, and unique insights. Users pay the sum of all underlying model completions rather than a single model price.