Bland AI Turbo is Bland's low-latency call-processing model designed for the fastest, most natural-sounding conversations in voice, SMS, and chat. It is positioned for enterprise voice agents and emphasizes very low latency and highly realistic, nuanced dialogue suitable for ultra-responsive calling experiences. Turbo supports private/self-hosted deployment options and strong data controls (encryption and on-customer infrastructure) to preserve data sovereignty and compliance. Key product capabilities include fast, private, brand-voice AI for calls/SMS/chat, custom model training from a company's recordings/transcripts, optional voice actor or voice cloning for brand alignment, per-conversation analytics (sentiment, call scoring), memory and integrations (Slack, HubSpot, CRM/ERP), and turnkey deployment support with Forward Deployed Engineers to tune voice, tone, prompts, rhythm, and guardrails. Bland offers two primary models: Base and Turbo. Base is the default, full feature set and is recommended for workflows that require integrations, script following, transfers, IVR, and custom tools. Turbo is optimized for speed and natural dialog and delivers lowest latency and very realistic conversations but has a reduced capability set: notable limitations include excluded or limited transfers, IVR navigation, and custom tool integrations. Bland's call processing uses three LLM roles (Navigational for flow decisions, Conversational for utterances, and Data Extraction for structured variable extraction). Pricing is primarily usage-based: voice (call time) is billed at $0.09 per minute (prorated to the second); there is an outbound minimum per dispatched call of $0.015 (billing update dated June 16, 2025) for short/failed connections; SMS is $0.02 per message. Public plan tiers listed include Start (Free) with small caps (example daily cap ~100 calls, concurrency ~10) and 1 voice clone; Build at $299/month with higher caps, concurrency ~50, and 5 voice clones; Scale at $499/month with larger caps, concurrency ~100, and 15 voice clones; and Enterprise (contact sales) with custom pricing, unlimited usage & concurrency, and dedicated terms. Bland's marketing materials emphasize scalability (marketing reference to up to ~1M concurrent calls). Developer and docs notes: the API allows selecting the model per call/agent (model field can be set to "base" or "turbo"); docs commonly recommend Base for many cases. The Create/Update Web Agent docs describe prompt best practices, webhook payloads (including analysis and extracted fields), voice selection, pathways, language settings, and other call configuration options. Sources include Bland product and docs pages and a Bland University module. Limitations and governance: Bland emphasizes guardrails to keep agents on-script and on-brand; because Turbo has a reduced capability set, some governance/tooling flows (transfers, IVR, custom tool integrations) may not be available when using Turbo. Bland markets self-hosting/private infrastructure options to help preserve data sovereignty and compliance. Sources referenced: https://www.bland.ai/turbo; https://docs.bland.ai/platform/billing; https://docs.bland.ai/api-v1/post/agents; https://university.bland.ai/modules/1/lesson-4 and related Bland docs and blog posts.