{"company":{"name":"AssemblyAI","slug":"assemblyai","website":"https://www.assemblyai.com","category":"ai-ml"},"question":"What has AssemblyAI shipped recently?","answer":"In the last 30 days, AssemblyAI shipped 16 tracked updates. The most recent was \"Best dictation APIs in 2026: 9 options compared for developers\" on 2026-10-07.","window":{"days":30,"updateCount":16,"returned":16},"generatedAt":"2026-10-08T10:38:40.150Z","updates":[{"title":"Best dictation APIs in 2026: 9 options compared for developers","summary":"AssemblyAI introduced its Dictation API, designed to return finished text (cleaned of fillers, self-corrections, and punctuated) instead of raw transcripts, addressing a gap in dictation workflows. The API runs on Universal-3.5 Pro at $0.62/hr and includes both verbatim and cleaned outputs per request. Competitors like Deepgram, ElevenLabs, and OpenAI return verbatim transcripts requiring manual cleanup.","date":"2026-10-07","dateIsEstimated":false,"signalType":null,"signalTypeLabel":null,"sourceUrl":"https://www.assemblyai.com/blog/best-dictation-apis","publisher":"assemblyai.com"},{"title":"How to choose the best speech-to-text API in 2026","summary":"AssemblyAI released a comprehensive guide to selecting the best speech-to-text API in 2026, emphasizing benchmarking on real-world audio rather than just WER. The post highlights AssemblyAI’s Universal-3.5 Pro and Universal-3.6 Pro Realtime models, comparing their accuracy, latency, and features against competitors like ElevenLabs, Gladia, and Deepgram.","date":"2026-10-07","dateIsEstimated":false,"signalType":null,"signalTypeLabel":null,"sourceUrl":"https://www.assemblyai.com/blog/how-to-choose-the-best-speech-to-text-api-for-your-product","publisher":"assemblyai.com"},{"title":"Voice AI infrastructure with unmatched price-performance","summary":"AssemblyAI unveiled a new cloud-based voice AI infrastructure platform emphasizing unmatched price-performance, predictable usage-based pricing, and zero infrastructure management. The offering includes 99.9% uptime SLA, multi-region redundancy, and SOC 2/GDPR compliance, targeting production workloads for developers and enterprises.","date":"2026-10-05","dateIsEstimated":true,"signalType":"product_launch","signalTypeLabel":"Launch","sourceUrl":"https://www.assemblyai.com/deployments/cloud","publisher":"assemblyai.com"},{"title":"How to evaluate a speech-to-text API for an education platform (2026)","summary":"AssemblyAI released a guide on evaluating speech-to-text APIs for education platforms, emphasizing student-speech accuracy, long-form audio economics, FERPA compliance, and accessibility requirements. The post highlights AssemblyAI’s pay-as-you-go pricing, EU data residency, and support for high-volume backfills as differentiators.","date":"2026-09-30","dateIsEstimated":false,"signalType":"content_marketing","signalTypeLabel":"Content","sourceUrl":"https://www.assemblyai.com/blog/how-to-evaluate-speech-to-text-api-education-platform","publisher":"assemblyai.com"},{"title":"The voice agent launch runbook: ramp, rollback, and on-call","summary":"AssemblyAI published a detailed runbook for operating its Voice Agent API post-launch, covering staged traffic ramps, rollback thresholds, severity definitions, and on-call ownership. The guidance addresses unique challenges of voice agents, such as conversational failures and heterogeneous traffic patterns, and introduces canary routing by conversation counts rather than time.","date":"2026-09-30","dateIsEstimated":false,"signalType":"feature_update","signalTypeLabel":"Feature","sourceUrl":"https://www.assemblyai.com/blog/voice-agent-launch-runbook","publisher":"assemblyai.com"},{"title":"When keyterm prompting backfires","summary":"AssemblyAI’s blog details reproducible cases where keyterm prompting worsens transcript accuracy by overriding correct words or over-prompting, particularly with similar-sounding terms. It introduces a pre-flight checklist and advocates for scoped, contextual prompting over flat keyterm lists, especially in multilingual deployments.","date":"2026-09-30","dateIsEstimated":false,"signalType":"content_marketing","signalTypeLabel":"Content","sourceUrl":"https://www.assemblyai.com/blog/when-keyterm-prompting-backfires","publisher":"assemblyai.com"},{"title":"Ambient scribes beyond healthcare: Building for veterinary and legal","summary":"AssemblyAI detailed how its ambient AI scribe technology, originally built for healthcare, adapts to veterinary and legal documentation needs. The post highlights key architectural differences, including species-specific terminology, multi-speaker diarization, and structured extraction requirements for legal depositions versus client meetings.","date":"2026-09-30","dateIsEstimated":false,"signalType":"feature_update","signalTypeLabel":"Feature","sourceUrl":"https://www.assemblyai.com/blog/ambient-scribes-veterinary-legal","publisher":"assemblyai.com"},{"title":"How to build clinical dictation on AssemblyAI","summary":"AssemblyAI published a technical guide distinguishing clinical dictation (short-clip) from ambient scribing, clarifying when to use the Sync API (verbatim transcript) versus the Dictation API (cleaned, steerable output). Medical Mode is not available on either endpoint, requiring use of Universal-3.5 Pro/3.6 Pro with domain: 'medical-v1' for clinical accuracy.","date":"2026-09-30","dateIsEstimated":false,"signalType":"feature_update","signalTypeLabel":"Feature","sourceUrl":"https://www.assemblyai.com/blog/build-clinical-dictation-assemblyai","publisher":"assemblyai.com"},{"title":"Your agent can take no for an answer","summary":"AssemblyAI released Universal-3.6 Pro Realtime, a major update to their streaming speech-to-text model, reducing word error rates by 16-52% across scenarios like noisy environments, background speech, and code-switching. The model now handles 32 languages with automatic detection, improves end-of-turn precision, and introduces voice focus to suppress competing talkers. It maintains the same latency and pricing ($0.45/hr) as Universal-3.5 Pro.","date":"2026-09-29","dateIsEstimated":false,"signalType":"feature_update","signalTypeLabel":"Feature","sourceUrl":"https://www.assemblyai.com/blog/universal-3-6-pro-realtime","publisher":"assemblyai.com"},{"title":"Universal-3.6 Pro Realtime: trained on real conversations, built for real voice agents","summary":"AssemblyAI released Universal-3.6 Pro Realtime, a speech-to-text model trained on tens of thousands of hours of real voice-agent and call-center conversations. It reduces errors on short replies by 45%, cuts background-speech leakage by 29%, and lowers language confusion by 74% compared to Universal-3.5 Pro Realtime. The model also adds support for 14 new languages, including Korean and Russian, and introduces entity-aware endpointing for better handling of digit strings and identifiers.","date":"2026-09-29","dateIsEstimated":false,"signalType":null,"signalTypeLabel":null,"sourceUrl":"https://www.assemblyai.com/blog/universal-3-6-pro-realtime-research","publisher":"assemblyai.com"},{"title":"Universal-3.5 Pro is now on OpenRouter","summary":"AssemblyAI announced Universal-3.5 Pro, its top-ranked speech-to-text model, is now available on OpenRouter’s transcription endpoint. The integration allows developers to access the model through a single API key and billing relationship, with pricing matching AssemblyAI’s published rates and no markup. Existing OpenAI SDK clients can use the model without code changes.","date":"2026-09-22","dateIsEstimated":false,"signalType":"integration","signalTypeLabel":"Integration","sourceUrl":"https://www.assemblyai.com/blog/aai-on-openrouter","publisher":"assemblyai.com"},{"title":"AssemblyAI adds Speaker Diarization for pre-recorded audio","summary":"AssemblyAI introduced Speaker Diarization for pre-recorded audio, enabling automatic labeling of individual speakers in transcripts. The feature splits transcripts into utterances per speaker with configurable constraints like exact speaker counts or expected ranges, improving multi-speaker accuracy. It is available via API and SDKs with detailed configuration options.","date":"2026-09-19","dateIsEstimated":true,"signalType":"feature_update","signalTypeLabel":"Feature","sourceUrl":"https://www.assemblyai.com/docs/pre-recorded-audio/label-speakers","publisher":"assemblyai.com"},{"title":"AssemblyAI expands LLM Gateway with new models, pricing, and EU server support","summary":"AssemblyAI’s LLM Gateway now lists multiple new models—including Qwen3.5 4B Fast, Claude Sonnet 4.6, and Google’s Gemini 2.5 Flash Lite—with detailed context lengths, supported parameters, and pricing tiers. The update also introduces EU server support via a dedicated endpoint and clarifies regional pricing differences.","date":"2026-09-15","dateIsEstimated":true,"signalType":"feature_update","signalTypeLabel":"Feature","sourceUrl":"https://www.assemblyai.com/docs/llm-gateway/api-reference/list-available-models","publisher":"assemblyai.com"},{"title":"Introducing the Dictation API: the first API built for dictation","summary":"AssemblyAI introduced the Dictation API, a new product designed to convert spoken utterances into clean, ready-to-use text by resolving self-corrections, removing filler words, and preserving user-intended phrasing. It runs on Universal-3.5 Pro across 19 languages at $0.62/hr and accepts audio in chunks for sub-second latency.","date":"2026-09-15","dateIsEstimated":false,"signalType":"product_launch","signalTypeLabel":"Launch","sourceUrl":"https://www.assemblyai.com/blog/dictation-api","publisher":"assemblyai.com"},{"title":"AssemblyAI launches LLM Gateway with multi-model routing, regional compliance, and post-processing","summary":"AssemblyAI introduced LLM Gateway, a unified API for accessing 25+ LLMs across providers like Anthropic, Google, OpenAI, and Qwen with automatic fallbacks and regional endpoints (US/EU). The service supports streamed responses, structured outputs, tool calling, and post-processing features like JSON repair, with compliance-focused regional routing and optional global model discounts.","date":"2026-09-11","dateIsEstimated":true,"signalType":"feature_update","signalTypeLabel":"Feature","sourceUrl":"https://www.assemblyai.com/docs/llm-gateway/quickstart","publisher":"assemblyai.com"},{"title":"AssemblyAI expands SDK support for Python and JavaScript","summary":"AssemblyAI now offers official, open-source SDKs for Python and JavaScript covering pre-recorded transcription, real-time streaming, and speech understanding. The SDKs handle authentication, polling, and retries, simplifying integration for developers.","date":"2026-09-11","dateIsEstimated":true,"signalType":"feature_update","signalTypeLabel":"Feature","sourceUrl":"https://www.assemblyai.com/docs/getting-started/sdks","publisher":"assemblyai.com"}],"attribution":{"source":"Spyingbee","url":"https://spyingbee.com/updates/assemblyai","citation":"Spyingbee, \"AssemblyAI updates\", https://spyingbee.com/updates/assemblyai (retrieved 2026-10-08)"}}