#model-routing

GPT-5.6 Luna fell 80% and Terra 20% on July 30, Sonnet 5's promo ends August 31, and OpenRouter list prices quietly mix standard with batch rates.

Qwen3.7 Flash listed at $0.03 per million input tokens on its base tier, 1M context and video input — but no published benchmarks. Where it fits in subagents.

Microsoft's MAI-Cyber-1-Flash claims 96% on CyberGym at half the cost. The routing thesis behind it matters more than the unverified benchmark.

Runway's Media Router sends one API call to its own models or to rivals, with hard USD price caps per second or per image and dry-run routing validation.

Claude Opus 5 launched July 24 at $5/$25 per million tokens — Anthropic's new state of the art on coding and knowledge work at half Fable 5's price.

AI cost control has split into four stackable levers: routing, planner/worker splits, token efficiency, and model choice. A 2026 playbook for cutting AI spend.

Cursor Router auto-classifies each coding request and picks the right model. Cursor claims up to 60% savings in A/B tests versus running everything on Opus.

Seedream 5.0 Pro joins Vercel AI Gateway alongside GPT-5.6 and Muse Spark 1.1. How to route image models for on-brand ad creative at scale in 2026.

Tencent's Hy3 shipped Apache 2.0 at a sub-300GB FP8 footprint. A routing guide: send agentic tool-use to Hy3 and repo-scale coding to a frontier model.

Fable 5 meters from July 7. Spend the remaining included window on durable plan artifacts that cheaper models can execute for months, not on routine drafts.

Claude Fable 5 was included through July 7 — now extended to July 12 — then moves to metered usage credits. How to sequence the window before the meter starts.

Run Claude Fable 5 as the planner and route execution to cheaper models in Hermes and OpenClaw. The exact config keys, cost math, and marketplace hardening.

Ten one-shot audit and upgrade jobs for Claude Fable 5, each drawn from Anthropic's own prompting guide, with the cheaper model that should execute the plan.

Claude Fable 5 costs $10/$50 per million tokens, double Opus 4.8. How individuals and small firms recoup that metered spend — with the honest counterexamples.

Claude Fable 5 pricing resolved July 18: Max and Team Premium keep 50% included access; Pro and Team Standard get $10/$50 usage credits from July 20 on.

Why AI services don't earn SaaS margins: inference COGS, the 50-60% gross-margin reality, model-routing levers, and hybrid pricing for agency work.

Uber burned a year's AI budget in four months; Microsoft cut Claude Code. A 4-gate playbook that right-sizes model spend and cuts AI bills 60 to 80 percent.

The Fable 5 export shutdown showed single-vendor AI can halt your business overnight. A four-step second-source playbook with open-weight failover backups.

NVIDIA shipped Nemotron 3 Ultra, a 550B open MoE reasoning model with weights, data and recipes under a permissive license. It runs fast but trails Kimi K2.6.

MiniMax M3 lands at 5-17x lower cost, but Opus 4.8 leads SWE-bench Pro and GPT-5.5 wins Terminal-Bench. A full three-way agentic coding routing matrix.

The LLM gateway is now critical AI infrastructure. Compare LiteLLM, Portkey, Cloudflare, Vercel, and OpenRouter on caching, routing, and build-vs-buy economics.

Gemini 3.5 Flash beats Claude Opus 4.8 on MCP-Atlas and Finance Agent at a third of the price — but a 61% hallucination rate complicates the routing call.

A FinOps playbook for cutting AI inference spend without quality loss: model routing, prompt and KV caching, batching, quantization, and unit-cost tracking.

Hands-on deep dive into Zed's AI coding — Parallel Agents, channels, threads, performance-first editor, model routing, and the workflows enabled.

Hands-on deep dive into Continue.dev — the open-source AI coding assistant, model routing, context providers, slash commands, IDE integrations.

Hands-on deep dive into Cursor 3 — Agents and Cloud Agents, Composer evolution, MCP integration, model routing, and what's new vs Cursor 2.

OpenRouter Fusion sends queries to multiple AI models, analyzes outputs, and fuses optimal results. Deep Research agents preferred Fusion to their own outputs.