PRICE CHANGES

Muse Spark 1.3 joins the board at $1.25 in / $4.25 out per MTok. Meta also sells it at $0.10 / $0.20 if you let them train on your data. We list the cash price; the other one is in the note.

Muse Wonder, Meta's first frontier model, joins the board at $4 in / $20 out per MTok. That's Opus 5.5 to the cent, cache reads included. If it burns through your limits faster, blame token count, not the rate.

Gemini 4 Argon joins the board at $2 in / $10 out per MTok, same as GPT-6 Sol. It doubles to $4/$20 on Dec 1, which is exactly Opus 5.5 pricing. Cache reads run 95% below input.

Claude Haiku 5.5 joins the board at $0.10 in / $0.50 out per MTok: 10x below the Haiku 4.5 it succeeds, level with GPT-6 Luna as the cheapest model we track. Past 100K prompt tokens: $0.50/$2.50.

GLM 5.3 Flash joins the board at $0.15 in / $0.50 out per MTok — 11.2% of all OpenRouter tokens last week, more than any western model we track except DeepSeek.

GLM 5.3 joins the board at $1.40 in / $4.40 out per MTok — the first Chinese open models we track, after a board that covered 38% of the traffic it claims to index.

GPT-6.1 Sol joins the board at $2.00 in / $10.00 out per MTok — same list price as GPT-6 Sol, but cache reads cost half: $0.20 down to $0.10/MTok.

Claude Sonnet 5.5 joins the board at $2.00 in / $10.00 out per MTok — identical list, cache and batch rates to the Sonnet 5 it succeeds. No price news is the price news.

Claude Opus 5.5 joins the board at $4.00 in / $20.00 out per MTok — the first cut inside the Opus line: 20% below the Opus 5 it succeeds, with cache reads down to $0.20/MTok.

GPT-6 Sol joins the board at $2.00 in / $10.00 out per MTok — GPT-6 ships tiered after all, and this one undercuts the GPT-5.6 Terra it sits beside on output.

GPT-6 Luna joins the board at $0.10 in / $0.50 out per MTok — the cheapest model we track, 100x below GPT-6 Astra and half the price of GPT-5.6 Luna.