--- title: Fish Audio Releases Drama 3, a 35B Open-Weight MoE... description: Fish Audio released Drama 3, a 35B-parameter mixture-of-experts text-to-speech model with about 3B active parameters per token, built on the Qwen3.5-35B-A3B... canonical: https://daily.dev/posts/fish-audio-releases-drama-3-a-35b-open-weight-moe-text-to-speech-model-in83k7spo twitter:card: summary_large_image twitter:site: @dailydotdev og:type: website og:site_name: daily.dev og:title: Fish Audio Releases Drama 3, a 35B Open-Weight MoE Text-to-Speech Model | daily.dev og:description: Fish Audio released Drama 3, a 35B-parameter mixture-of-experts text-to-speech model with about 3B active parameters per token, built on the Qwen3.5-35B-A3B... og:url: https://daily.dev/posts/fish-audio-releases-drama-3-a-35b-open-weight-moe-text-to-speech-model-in83k7spo og:image: https://api.daily.dev/og/posts/in83k7SPO.png og:image:alt: Fish Audio Releases Drama 3, a 35B Open-Weight MoE Text-to-Speech Model og:image:width: 1200 og:image:height: 630 og:locale: en --- > ## Documentation Index > Fetch the complete documentation index at: https://daily.dev/llms.txt > Use this file to discover all available pages before exploring further. # Fish Audio Releases Drama 3, a 35B Open-Weight MoE Text-to-Speech Model **[Machine Learning News](https://daily.dev/sources/mlnews)** · 5 min read · 0 upvotes · 0 comments ## Summary Fish Audio released Drama 3, a 35B-parameter mixture-of-experts text-to-speech model with about 3B active parameters per token, built on the Qwen3.5-35B-A3B backbone. It accepts plain-language acting directions and emotional cues, supports 83 languages, and can render up to 16 speakers in a single API request. The model scores 93.2% on Artificial Analysis Pronunciation Robustness (best-in-class for Fish Audio) but trails ElevenLabs' Eleven v4 Turbo on overall quality Elo. Self-hosting requires at least 48GB of NVIDIA GPU memory in FP8 (no CPU, Apple MPS, or AMD ROCm support); official benchmarks run on an H200. Weights are available on Hugging Face under the Fish Audio Community License v1.0, which allows free commercial use under $500,000 in annual revenue. The hosted API prices at $50 per 1M characters, discounted to $10 until November 12, 2026. ## Full article daily.dev links to this article rather than hosting it. Read it at the original source: ## Questions this post answers ### What hardware do I need to self-host Fish Audio Drama 3? Drama 3 requires at least one NVIDIA GPU with 48GB or more of memory running in FP8 precision; CPU, Apple MPS, and AMD ROCm are not supported. MoE experts ship in FP8 while remaining weights use bf16. Official streaming benchmarks were measured on a single NVIDIA H200, achieving 88.9ms median time to first audio at one stream. _Developers evaluating GPU requirements for self-hosted TTS models can track releases like Drama 3 on daily.dev._ ### How much does the Fish Audio Drama 3 API cost and is it free for commercial use? The Drama 3 API lists at $50 per 1 million characters, with a discounted launch price of $10 per 1 million characters through November 12, 2026. Under the Fish Audio Community License v1.0, commercial use is free for companies under $500,000 in annual revenue; hosting Drama 3 as a third-party API requires a separate license. _Teams weighing licensing costs for voice AI tools can follow pricing changes like this on daily.dev._ ### How does Fish Audio Drama 3 compare to ElevenLabs on text-to-speech quality? Drama 3 Preview scores an Elo of 1242 on the Artificial Analysis quality leaderboard, ahead of its predecessor S2.1 Pro at 1190, but behind ElevenLabs Eleven v4 Turbo at 1328. On pronunciation robustness specifically, Drama 3 leads with 93.2% overall, scoring 95.9% on brands and technical terms but only 86.3% on exact sequences like codes and emails. _Developers choosing between TTS providers can compare benchmarks like these on daily.dev._ ## Similar posts on daily.dev - [ElevenLabs v3 Text-to-Speech on DigitalOcean Inference](https://daily.dev/posts/elevenlabs-v3-text-to-speech-on-digitalocean-inference-qekzeoxov) · DigitalOcean Community · 0 upvotes · 0 comments --- Tags: [#open-source](https://daily.dev/tags/open-source), [#mixture-of-experts](https://daily.dev/tags/mixture-of-experts), [#text-to-speech](https://daily.dev/tags/text-to-speech) [View this post on daily.dev](https://daily.dev/posts/fish-audio-releases-drama-3-a-35b-open-weight-moe-text-to-speech-model-in83k7spo) ```json {"@context":"https://schema.org","@graph":[{"@type":"Organization","@id":"https://daily.dev/#organization","name":"daily.dev","url":"https://daily.dev","logo":{"@type":"ImageObject","url":"https://daily.dev/apple-touch-icon.png","width":180,"height":180},"sameAs":["https://twitter.com/dailydotdev","https://github.com/dailydotdev","https://www.linkedin.com/company/daily-dev-ltd"]},{"@type":"WebSite","@id":"https://daily.dev/#website","url":"https://daily.dev","name":"daily.dev","publisher":{"@id":"https://daily.dev/#organization"},"potentialAction":{"@type":"SearchAction","target":{"@type":"EntryPoint","urlTemplate":"https://daily.dev/search?q={search_term_string}"},"query-input":"required name=search_term_string"}}]} {"@context":"https://schema.org","@type":"TechArticle","headline":"Fish Audio Releases Drama 3, a 35B Open-Weight MoE Text-to-Speech Model","url":"https://daily.dev/posts/fish-audio-releases-drama-3-a-35b-open-weight-moe-text-to-speech-model-in83k7spo","mainEntityOfPage":{"@type":"WebPage","@id":"https://daily.dev/posts/fish-audio-releases-drama-3-a-35b-open-weight-moe-text-to-speech-model-in83k7spo"},"datePublished":"2026-10-14T23:26:38.490Z","dateModified":"2026-10-15T08:39:54.661Z","description":"Fish Audio released Drama 3, a 35B-parameter mixture-of-experts text-to-speech model with about 3B active parameters per token, built on the Qwen3.5-35B-A3B...","image":"https://media.daily.dev/image/upload/f_auto,q_auto/v1/posts/a80db2b41d22188d2cf40853da57a342?_a=AQAEuop","thumbnailUrl":"https://media.daily.dev/image/upload/f_auto,q_auto/v1/posts/a80db2b41d22188d2cf40853da57a342?_a=AQAEuop","isAccessibleForFree":true,"articleSection":"Machine Learning News","inLanguage":"en","publisher":{"@type":"Organization","name":"daily.dev","url":"https://daily.dev","logo":{"@type":"ImageObject","url":"https://daily.dev/apple-touch-icon.png","width":180,"height":180}},"author":{"@type":"Organization","name":"Machine Learning News","logo":"https://media.daily.dev/image/upload/s--0PoEtdkd--/f_auto/v1750409401/squads/marktechpost?_a=BAMClqZW0","url":"https://daily.dev/squads/mlnews"},"commentCount":0,"discussionUrl":"https://daily.dev/posts/fish-audio-releases-drama-3-a-35b-open-weight-moe-text-to-speech-model-in83k7spo","interactionStatistic":[{"@type":"InteractionCounter","interactionType":{"@type":"LikeAction"},"userInteractionCount":0},{"@type":"InteractionCounter","interactionType":{"@type":"CommentAction"},"userInteractionCount":0}],"keywords":"open-source,mixture-of-experts,text-to-speech","timeRequired":"PT5M"} {"@context":"https://schema.org","@type":"BreadcrumbList","itemListElement":[{"@type":"ListItem","position":1,"name":"Home","item":"https://daily.dev"},{"@type":"ListItem","position":2,"name":"Machine Learning News","item":"https://daily.dev/squads/mlnews"},{"@type":"ListItem","position":3,"name":"Fish Audio Releases Drama 3, a 35B Open-Weight MoE Text-to-Speech Model"}]} {"@context":"https://schema.org","@type":"FAQPage","@id":"https://daily.dev/posts/fish-audio-releases-drama-3-a-35b-open-weight-moe-text-to-speech-model-in83k7spo#faq","mainEntity":[{"@type":"Question","name":"What hardware do I need to self-host Fish Audio Drama 3?","acceptedAnswer":{"@type":"Answer","text":"Drama 3 requires at least one NVIDIA GPU with 48GB or more of memory running in FP8 precision; CPU, Apple MPS, and AMD ROCm are not supported. MoE experts ship in FP8 while remaining weights use bf16. Official streaming benchmarks were measured on a single NVIDIA H200, achieving 88.9ms median time to first audio at one stream. Developers evaluating GPU requirements for self-hosted TTS models can track releases like Drama 3 on daily.dev."}},{"@type":"Question","name":"How much does the Fish Audio Drama 3 API cost and is it free for commercial use?","acceptedAnswer":{"@type":"Answer","text":"The Drama 3 API lists at $50 per 1 million characters, with a discounted launch price of $10 per 1 million characters through November 12, 2026. Under the Fish Audio Community License v1.0, commercial use is free for companies under $500,000 in annual revenue; hosting Drama 3 as a third-party API requires a separate license. Teams weighing licensing costs for voice AI tools can follow pricing changes like this on daily.dev."}},{"@type":"Question","name":"How does Fish Audio Drama 3 compare to ElevenLabs on text-to-speech quality?","acceptedAnswer":{"@type":"Answer","text":"Drama 3 Preview scores an Elo of 1242 on the Artificial Analysis quality leaderboard, ahead of its predecessor S2.1 Pro at 1190, but behind ElevenLabs Eleven v4 Turbo at 1328. On pronunciation robustness specifically, Drama 3 leads with 93.2% overall, scoring 95.9% on brands and technical terms but only 86.3% on exact sequences like codes and emails. Developers choosing between TTS providers can compare benchmarks like these on daily.dev."}}]} ```