Deepgram Enhanced: model profile

Overview

Deepgram introduced Enhanced in May 2022, claiming 19% higher relative accuracy than the previous model and improved handling of long-tail vocabulary. Deepgram positioned Enhanced as the middle tier between Base and custom-trained models: better out-of-the-box quality than Base, without the cost and effort of full custom model training. In current Deepgram documentation, Enhanced remains available with variants such as general, meeting, phonecall, and finance, while Deepgram's 2025 to 2026 product narrative centers on newer speech products such as Nova-3 and Flux.

Enhanced is also exposed through Future AGI's Agent Command Center, a gateway and control-plane layer that sits between an application and model providers. Future AGI's public model catalog lists routable model IDs including deepgram/enhanced, deepgram/enhanced-general, deepgram/enhanced-phonecall, deepgram/enhanced-meeting, and deepgram/enhanced-finance, advertised as callable through Agent Command Center with unified observability, caching, fallback, and "15 routing strategies including cost-optimized fallback."

Capabilities and features

Vendor-reported capabilities from Deepgram's 2022 launch and product materials:

• 19% higher relative accuracy than the previous model, better word recognition, and stronger long-tail vocabulary handling.
• An "increased effective vocabulary" that handles infrequent and uncommon words significantly better than lower tiers, positioned for teams that need stronger recognition of domain terms without committing to a custom-training program.
• Variants: general, meeting (beta), phonecall, finance (beta).

Enhanced uses Deepgram's standard STT interfaces rather than a bespoke API. Pre-recorded STT uses the /v1/listen family and accepts audio and video inputs, with JSON responses for transcription results or a request ID for callback and asynchronous paths. Streaming uses WebSockets and supports standard streaming controls and transcript options such as sample rate, smart formatting, diarization, interim and final results, and supported audio encodings. Deepgram's docs state support for 100+ audio formats and encodings overall.

Deepgram advised customers that model upgrades can change transcript outputs and recommended side-by-side testing with features such as keywords before production rollout.

Language support

English availability was announced at launch in May 2022. Enhanced support expanded to German in October 2022. No further language coverage for Enhanced is stated in the sources reviewed.

Performance and benchmarks

Deepgram does not publish model-specific WER benchmarks, named benchmark datasets, or competitor tables for Enhanced in the sources reviewed. By 2025 to 2026, Deepgram's public benchmark language moved to newer models such as Nova-2, Nova-3, and Flux.

Third-party evaluation context: independent research shows that speech systems from top providers, including Deepgram, can fail on short high-stakes utterances, non-native accented English, aphasia speech, and other off-benchmark conditions. These studies are cross-provider and not Enhanced-specific. Deepgram publishes multiple pieces advising customers to evaluate against production audio rather than headline WER.

Deepgram model context

Latency and throughput

Deepgram does not disclose latency-to-first-token or latency distributions for Enhanced specifically. Vendor-reported platform-level figures from 2022 materials: use-case models able to transcribe one hour of pre-recorded audio in about 30 seconds, and real-time streaming with less than 300 ms lag. These are family and platform-level claims, not Enhanced-only measurements.

Current Deepgram rate-limit docs still list Enhanced. On starter-like limits visible in the docs, Enhanced is shown at up to 50 concurrent pre-recorded requests and up to 150 concurrent streaming requests in North America/Europe tables, rising on Growth and Enterprise plans. These figures are service limits, not intrinsic model throughput measurements.

Deployment and integrations

Deepgram exposes Enhanced through its hosted API (pre-recorded /v1/listen and WebSocket streaming). Deepgram publishes official developer SDKs and playground tooling for STT, TTS, and voice APIs, including the API Playground, JavaScript SDK, and STT getting-started guides for pre-recorded and live transcription. Deepgram's trust and self-hosting materials state that self-hosted deployments typically keep audio and transcripts inside customer infrastructure, apart from license validation and usage reporting, with hardened systems, RBAC, encryption, SOC 2 controls, HIPAA-compatible offerings for enterprise customers, and GDPR readiness with EU data residency.

Future AGI Agent Command Center

Future AGI's Agent Command Center is a gateway and control-plane layer between the client application and upstream providers. Each request passes through a fixed-order plugin chain: IP ACL, authentication, RBAC, cache lookup, budget checks, guardrails, tool policy, validation, and rate limiting before the provider call; the response then continues through cost tracking and logging. Cache hits can short-circuit the provider call entirely.

Future AGI's public Deepgram model pages state that calling models such as deepgram/enhanced-general through Agent Command Center returns metadata on x-agentcc-* headers including provider, cost, latency, cache hit, and request ID. The catalog advertises 36 Deepgram models, including Enhanced variants, as routable through Agent Command Center.

Documentation mismatch: Agent Command Center's provider docs enumerate OpenAI, Anthropic, Gemini, Bedrock, Azure, Cohere, Groq, Mistral, Together, Fireworks, DeepInfra, Perplexity, Cerebras, xAI, OpenRouter, Hugging Face, Anyscale, Replicate, and several self-hosted backends, but do not explicitly mention Deepgram. The primary API reference frames /v1/audio/transcriptions in Whisper/OpenAI-compatible terms. The public sources reviewed do not resolve this discrepancy.

Gateway characteristics from Future AGI's public docs and README:

• Observability: every request is logged with request ID, trace ID, session ID, model requested vs model actually used, provider, token counts, cost, latency, cache status, guardrail results, and fallback/error events. Metrics export to Prometheus and traces to OpenTelemetry; users can propagate their own trace IDs via request headers.
• Caching: server-side exact and semantic strategies with configurable TTL, namespaces, LRU eviction, force-refresh, and Cache-Control: no-store. The open-source README states support for 6 exact-cache backends (mem, redis, disk, s3, gcs, azblob) and 4 semantic-cache backends (mem, pinecone, qdrant, weaviate).
• APIs and SDKs: OpenAI-compatible endpoints including /v1/chat/completions, /v1/embeddings, /v1/audio/transcriptions, /v1/audio/speech, /v1/images/generations, and /v1/responses. Future AGI publishes a client SDK repository for Python and TypeScript; the control-plane/gateway service lives in the platform monorepo.
• Deployment: hosted gateway, Docker, and Go-binary self-hosting from a config.yaml, intended to keep requests inside the customer's infrastructure. The README states clustering/HA uses Raft-based clustering; official Kubernetes manifests and Helm charts were "coming soon" at the time of the README snapshot reviewed.
• Security: virtual API keys, encrypted storage of provider credentials, RBAC, rate limits, budget controls, IP allowlists, and append-only audit logging. Prompts and completions are not stored by default; caching is opt-in and configurable per organization.

Routing strategies

Future AGI's public README names 15 routing and reliability strategies: roundrobin, latency, costopt, adaptive, complexity, conditional, providerlock, accessgroups, race/hedged, mirror/shadow, modelfallback, failover, circuitbreaker, retry, healthmonitor. The formal routing docs fully explain only a subset.

Pricing

Not publicly disclosed for Enhanced in the sources reviewed. Future AGI's public Deepgram calculator pages mark pricing, context window, and benchmarks for Enhanced variants as not currently public or pending.

Development and ownership

Deepgram was founded in 2015. Its official company story states the company emerged from machine-learning work for waveform analysis in a dark-matter detector in China, and that Scott Stephenson, identified by Deepgram as CEO and co-founder, later explored deep learning for audio analysis at the University of Michigan before building Deepgram around end-to-end deep learning. Deepgram's official author bio states Stephenson earned a PhD in particle physics from the University of Michigan and left a postdoctoral research role to found the company.

In January 2026, Deepgram raised a $130M Series C at a $1.3B valuation and acquired OfOne. Its Series C press release cites US patent 12,499,875, "Deep Learning Internal State Index-Based Search and Classification." USPTO patentee indexes also show Deepgram assignee entries for "Hardware efficient automatic speech recognition" (granted June 2025) and "End-to-end automatic speech recognition with transformer" (granted August 2025). These patents do not provide a public model card for Enhanced.

Future AGI is an open-source, end-to-end AI agent engineering platform spanning tracing, evaluation, optimization, protection, and gateway functions. Public profiles identify Nikhil Pareek as Founder and CEO, and Charu Gupta as co-founder; public sources indicate Future AGI was founded in 2024, though that founding year is not clearly stated on the official homepage reviewed.

Release history

Deepgram's current marketing and changelog emphasis favors Nova-3, Flux, multilingual expansion, medical specialization, and voice-agent stacks rather than further public narrative around Enhanced; Enhanced is still supported but is no longer Deepgram's flagship public innovation story.