Skip to content
Back
Pricing briefing
MMindsDB
MindsDB
PACKAGING CHANGE

MindsDB added a new Claude Haiku model to its rate card

By , Co-founder, PricingSaaS

Published

MindsDB added Claude Haiku 5.5 pricing to its token rate card.

MindsDB’s pricing page · Change detected

What changed

MindsDB added Claude Haiku 5.5 to its model rate card at $0.10 per 1M input tokens and $0.50 per 1M output tokens, with cached input at $0.01 and cache writes at $0.125 per 1M tokens, per MindsDB's pricing page as of Nov 7, 2026. For prompts over 100K tokens, the same rate card steps those rates up to $0.50 input and $2.50 output per 1M tokens, also as of Nov 7, 2026.

MindsDB also added an unversioned "Claude Haiku" alias row at the same base rates with no long-prompt surcharge, and left Haiku 4.5 unchanged at $1.00 input and $5.00 output per 1M tokens, per MindsDB's pricing page as of Nov 7, 2026.

Why it matters

Pulse tracked 18 changes to MindsDB's rate card in the last 90 days, averaging roughly one every 9 days. That cadence is what pricing a model marketplace, rather than a single product, looks like: each new foundation-model release becomes its own priced row, and existing rows can move too — Pulse recorded MindsDB raising its MindsHub Blaze rates in late September 2026.

For a product leader reselling or bundling third-party AI, the lesson is to price by model tier and time-box any rate guarantee you give customers — don't assume today's per-token economics hold for more than a few weeks when your own vendor resells someone else's compute. Watch whether MindsDB settles its rate card or keeps stacking new model rows: 28 of MindsDB's 35 pricing-page comparisons were real changes since Pulse began tracking on Mar 9, 2026, which points to a catalog still being built out rather than settled.

In context

Per Pulse's tracking, MindsDB has added a new model row every few weeks this year: Kimi K3.1 Pro and DeepSeek V4.1 Flash on Oct 27, and GPT-6.1 Terra just nine days before this change, on Oct 29 — so this new Claude Haiku row is part of a catalog that grows roughly every 9 days rather than a one-off repricing. Pulse tracked 328 credits-and-usage-billing changes across 224 companies in the last 180 days, 12.1% of the 2705 changes it tracked in that window, with Pika cutting its own credit caps on Nov 6 and Mintlify renaming a usage feature on Nov 5. The sharper read: MindsDB left Haiku 4.5 at $1.00 input and $5.00 output per 1M tokens and listed Haiku 5.5 at a tenth of that, so the spread between model rows, not the sticker price of any single row, is where the packaging decision sits.

Common questions

How should we price our own AI features if we resell models like this one?

Price by the specific model row you route traffic to, not a blanket "AI tier": Haiku 5.5's input and output rates differ sharply from Haiku 4.5's, and MindsDB can move that gap again whenever it lists a new model — plan to re-price your own usage-based line each time the upstream rate changes, not on a fixed annual cycle.

What should we check before routing long-context prompts to the new model?

Check your own average prompt length against the 100K-token line: MindsDB's rate card charges a higher rate once a single prompt crosses it, per MindsDB's pricing page as of Nov 7, 2026, so a workload with many long-context calls won't see the full headline rate on every request.

Sources