Every guide: ranked picks with struck-through official prices, a budget/balanced/quality decision matrix, honest methodology, FAQ — all from the live price table.
The "Pick for me" wizard now links each recommendation to its full guide.
2026-08-31
Model picker — choose the right model in three questions
/models now has a "Let Bridge pick for you" wizard: workload × volume × priority → a recommendation with real prices and a monthly cost estimate.
Every model carries a quality band (Flagship / Value / High-volume / Embedding) and a one-line "Best for" fit statement — on cards, the table and every model page.
New use-case filters (coding, reasoning, long-context & agents, vision, Chinese, high volume, embeddings) and a sortable table — by price, monthly cost, context, latency or savings.
New Est. /mo column: cost for a typical 10M-token/month workload (30% cache-hit), computed from the live price table.
2026-08-31
Catalog v2.1 — second supplier rate card
Pricing re-based on a second supplier's rate card (suggested catalog price, incl. tax): DeepSeek V4 Pro down to $0.28/$0.82 (78% below official peak), DeepSeek V4 Flash $0.14/$0.28 (67%), GLM-5.2 $0.56/$1.41 (50%), MiniMax M2.7 $0.28/$0.59.
Three new models: Kimi K2.6 (256K long-context generalist), Kimi K2.7 Code (coding specialist) and MiniMax M3 — all at $0.56/$1.13 or below.
Catalog is now 12 models; every price is live from the DB price table and reflected across the models page, model pages, blog and cost calculator.
Catalog expanded to 9 models: added GLM-5.3, Qwen3.7-Max and MiniMax M2.7; DeepSeek V4 Flash is now public.
Pricing re-anchored to our supplier channel rate card: official list prices are struck through next to Bridge prices on the models page and every model page (e.g. DeepSeek V4 Pro $0.57/$1.71 vs official $1.27/$3.80 — 55% off).
Model hub upgraded: vendor/type filters, latency chips, Official-channel badges, Featured and Best-value sections.