{"ok":true,"source":"tensorfeed.ai","lastUpdated":"2026-10-19","count":33,"models":[{"id":"midjourney-v8.2","name":"Midjourney V8.2","provider":"Midjourney","modality":"image","pricingUnit":"per_subscription_month","pricingAmount":10,"pricingNote":"Subscription only: Basic $10, Standard $30, Pro $60, Mega $120 per month (20% off annual). Extra fast GPU time $4/hr. No official API; access via the web app or Discord.","released":"2026-07-24","apiAvailable":false,"url":"https://docs.midjourney.com/hc/en-us/articles/32199405667853-Version","notes":"Default Midjourney model since July 24, 2026, replacing V8.1 (default from June 10) and V7. Ships a new Edit Model that replaces Omni Reference, Character Reference, and Retexture. Third-party \"Midjourney APIs\" are unofficial wrappers.","maxOutput":"2048px HD images","features":["personalization","style references","edit model","HD mode"]},{"id":"flux-1.1-pro-ultra","name":"FLUX 1.1 Pro Ultra","provider":"Black Forest Labs","modality":"image","pricingUnit":"per_image","pricingAmount":0.06,"pricingNote":"$0.06 per image (6 credits) via the Black Forest Labs API, same rate for Raw mode.","released":"2024-11-04","apiAvailable":true,"url":"https://blackforestlabs.ai/flux-1-1-ultra/","notes":"Previous-generation FLUX tier; BFL now recommends FLUX.2 for new projects. 4MP output without an upscaler, plus Raw mode for candid photography.","maxOutput":"4MP (2048x2048)","features":["photorealism","long prompts","raw mode"]},{"id":"flux-1.1-pro","name":"FLUX 1.1 Pro","provider":"Black Forest Labs","modality":"image","pricingUnit":"per_image","pricingAmount":0.04,"pricingNote":"$0.04 per image (4 credits) via the Black Forest Labs API. Replicate lists the same $0.04 per output image.","released":"2024-10-01","apiAvailable":true,"url":"https://docs.bfl.ai/flux_models/flux_1_1_pro","notes":"Previous-generation production workhorse, still served on the BFL API. BFL recommends FLUX.2 [pro] (from $0.03 per image, megapixel-based) for new text-to-image work.","maxOutput":"2MP (1408x1408)","features":["fast","json mode","image-to-image"]},{"id":"flux-1-schnell","name":"FLUX 1 Schnell","provider":"Black Forest Labs","modality":"image","pricingUnit":"per_image","pricingAmount":0.003,"pricingNote":"About $0.003 per image on Replicate (roughly 333 images per $1). Apache-2.0 licensed; free to self-host on a GPU.","released":"2024-08-01","apiAvailable":true,"url":"https://huggingface.co/black-forest-labs/FLUX.1-schnell","notes":"Open-weights distilled FLUX. 1-4 step generation. One of the cheapest hosted image models.","maxOutput":"1MP (1024x1024)","features":["open weights","apache-2.0","1-4 step"]},{"id":"gpt-image-2","name":"GPT Image 2","provider":"OpenAI","modality":"image","pricingUnit":"per_image","pricingAmount":0.053,"pricingNote":"Token-priced: $5 per 1M text input, $8 per 1M image input, $30 per 1M image output tokens. Batch API is half price ($15 per 1M image output). The per-image figure here is an earlier OpenAI estimate for medium quality at 1024x1024 ($0.006 low, $0.053 medium, $0.211 high); the pricing page now publishes token rates only and points at its cost calculator.","released":"2026-04-21","apiAvailable":true,"url":"https://developers.openai.com/api/docs/models/gpt-image-2","notes":"Replaced DALL-E 3, which OpenAI removed from the API on May 12, 2026. GPT Image 2.5 Sunburst and Flare (snapshot 2026-09-08) carry identical token rates and are now the named replacements on every retiring gpt-image notice. gpt-image-1 ($5 / $10 / $40) shuts down October 23, 2026; gpt-image-1.5 ($32 image output) and gpt-image-1-mini ($2 / $2.50 / $8, the cheapest image rate OpenAI sells) both go on December 1, 2026.","maxOutput":"3840px max edge (8.3MP)","features":["flexible sizes","image editing","transparent background","batch"]},{"id":"gemini-nano-banana-2.1","name":"Nano Banana 2.1","provider":"Google","modality":"image","pricingUnit":"per_image","pricingAmount":0.0336,"pricingNote":"Gemini API: $30 per 1M image output tokens, which Google equates to $0.0336 (1K), $0.0504 (2K) and $0.113 (4K) per image, plus $1.50 per 1M text, image or video input tokens and $7.50 per 1M text and thinking output tokens. Batch is half price.","released":"2026-10-06","apiAvailable":true,"url":"https://ai.google.dev/gemini-api/docs/models/gemini-nano-banana-2.1","notes":"Google's newest image model and now the replacement it names on the retiring Imagen 4 and Gemini 3.1 Flash Image Preview rows. Half the per-image price of Gemini 3.1 Flash Image at 1K and a quarter of Gemini 3 Pro Image, with configurable thinking (minimal, medium default, high). Takes text, image, video and PDF in; 131K input and 32K output tokens. No function calling, caching, structured outputs or Live API support.","maxOutput":"4K (1K default, 2K and 4K options)","features":["thinking","search grounding","image generation","batch"]},{"id":"gemini-3.1-flash-lite-image","name":"Gemini 3.1 Flash-Lite Image","provider":"Google","modality":"image","pricingUnit":"per_image","pricingAmount":0.0336,"pricingNote":"Gemini API: $30 per 1M image output tokens, stated as $0.0336 per 1K-resolution image, plus $0.25 per 1M text, image or video input tokens and $1.50 per 1M text and thinking output tokens. The cheapest input rate of any Google image model.","released":"2026-06-30","apiAvailable":true,"url":"https://ai.google.dev/gemini-api/docs/pricing","notes":"The budget image tier and the model Google names as the replacement for gemini-2.5-flash-image, whose Gemini API shutdown moved out to March 15, 2027. Same image output rate as Nano Banana 2.1 but a sixth of the text output rate, so it is the cheaper pick when prompts are long and resolution demands are modest.","maxOutput":"1K resolution images","features":["search grounding","image generation","batch"]},{"id":"gemini-3.1-flash-image","name":"Nano Banana 2 (Gemini 3.1 Flash Image)","provider":"Google","modality":"image","pricingUnit":"per_image","pricingAmount":0.067,"pricingNote":"Gemini API: $60 per 1M image output tokens, which Google equates to $0.045 (0.5K), $0.067 (1K), $0.101 (2K), and $0.151 (4K) per image. Batch is half price.","released":"2026-05-28","apiAvailable":true,"url":"https://ai.google.dev/gemini-api/docs/models/gemini-3.1-flash-image","notes":"GA version of the February 2026 preview. Imagen 4 was discontinued on Vertex AI June 30, 2026 and shut down on the Gemini API August 17, 2026, but Google's deprecations table now routes those users to Nano Banana 2.1 rather than here, and lists Nano Banana 2.1 as this model's own replacement too. Still GA with no shutdown date, and still twice the per-image price of the newer model. Supports conversational editing and video-to-image generation.","maxOutput":"4K (4096x4096)","features":["conversational editing","search grounding","video-to-image","batch"]},{"id":"gemini-3-pro-image","name":"Nano Banana Pro (Gemini 3 Pro Image)","provider":"Google","modality":"image","pricingUnit":"per_image","pricingAmount":0.134,"pricingNote":"Gemini API: $120 per 1M image output tokens, equivalent to $0.134 per 1K or 2K image and $0.24 per 4K image. Batch and Flex tiers are half price.","released":"2026-05-28","apiAvailable":true,"url":"https://ai.google.dev/gemini-api/docs/models/gemini-3-pro-image","notes":"GA version of the November 2025 preview and Google's highest-fidelity image model, with thinking before rendering. Google Cloud lists legible text rendering, multi-turn editing, and up to 14 reference inputs.","maxOutput":"4K","features":["thinking","text rendering","search grounding","multi-image reference"]},{"id":"recraft-v3","name":"Recraft v3","provider":"Recraft","modality":"image","pricingUnit":"per_image","pricingAmount":0.04,"pricingNote":"$0.04 per raster image, $0.08 per vector image, via prepaid API units (1,000 units = $1).","released":"2024-10-30","apiAvailable":true,"url":"https://www.recraft.ai","notes":"Strong on-image typography with vector output. Superseded by Recraft V4 ($0.04) and V4.1 ($0.035 per raster image), but V3 remains on the API and powers its editing endpoints.","maxOutput":"2048x2048","features":["typography","vector output","brand styles"]},{"id":"ideogram-2","name":"Ideogram 2.0","provider":"Ideogram","modality":"image","pricingUnit":"per_image","pricingAmount":0.08,"pricingNote":"$0.08 per image (2.0 Default) via API; 2.0 Turbo $0.05.","released":"2024-08-22","apiAvailable":true,"url":"https://ideogram.ai","notes":"Legacy model that is still billable on the API. Ideogram 4.0 (June 2026) is the current flagship at lower per-image prices.","maxOutput":"2048x2048","features":["typography","design styles","image-to-image"]},{"id":"ideogram-4.0","name":"Ideogram 4.0","provider":"Ideogram","modality":"image","pricingUnit":"per_image","pricingAmount":0.06,"pricingNote":"$0.06 per image (Default) via API; Turbo $0.03, Quality $0.10. Commercial self-hosting of the open weights requires a paid license.","released":"2026-06-03","apiAvailable":true,"url":"https://ideogram.ai/models/4.0/","notes":"Ideogram's first open-weight model: a 9.3B parameter diffusion transformer trained on structured JSON prompts, with bounding-box layout control and multilingual text rendering. Weights on Hugging Face (fp8 and nf4 checkpoints).","maxOutput":"native 2K","features":["open weights","text rendering","layout control","transparent background"]},{"id":"veo-3.1","name":"Veo 3.1","provider":"Google","modality":"video","pricingUnit":"per_second_video","pricingAmount":0.4,"pricingNote":"$0.40 per second with audio at 720p or 1080p, $0.60 at 4K (Gemini API and Google Cloud). Veo 3.1 Fast $0.10 to $0.30, Veo 3.1 Lite $0.05 to $0.08. Video-only output on Google Cloud is $0.20.","released":"2025-10-15","apiAvailable":true,"url":"https://docs.cloud.google.com/gemini-enterprise-agent-platform/models/veo/3-1-generate","notes":"Google's current Veo line with native synchronized audio. Replaces Veo 3, retired on both the Gemini API and Google Cloud on June 30, 2026. GA on Google Cloud (Gemini Enterprise Agent Platform) since November 17, 2025, but the three Gemini API preview ids (veo-3.1-generate-preview, -fast- and -lite-) shut down on October 22, 2026 with gemini-omni-1.1-flash named as the replacement. After that date Veo is a Google Cloud product and Gemini API callers go through Gemini Omni instead.","maxOutput":"8s @ 4K 24fps with audio","features":["native audio","first and last frame","extend","reference images"]},{"id":"kling-3.0","name":"Kling 3.0","provider":"Kuaishou","modality":"video","pricingUnit":"per_second_video","pricingAmount":0.084,"pricingNote":"Official API: $0.084/s at 720p, $0.112/s at 1080p, $0.42/s at 4K without audio; with native audio $0.126/s (720p) and $0.168/s (1080p). Kling 3.0 Turbo with audio $0.112 to $0.14/s. Billed in units (1 unit = $0.14).","released":"2026-02-05","apiAvailable":true,"url":"https://kling.ai/document-api/pricing/base/video","notes":"Kuaishou flagship with native multilingual audio, multi-shot storyboards, and a 3.0 Omni variant for reference-driven consistency. Kling 2.0 Master, 2.1, and older video models left the API on September 15, 2026, so 3.0 and its Turbo and Omni variants are the only Kling models still callable.","maxOutput":"15s @ 4K with audio","features":["native audio","multi-shot","image-to-video","motion control"]},{"id":"happyhorse-1.0","name":"HappyHorse 1.0","provider":"Alibaba","modality":"video","pricingUnit":"per_second_video","pricingAmount":0.14,"pricingNote":"$0.14 per second at 720P, $0.24 at 1080P (Model Studio, International). HappyHorse 1.1 lists $0.07 (480P), $0.14 (720P), and $0.18 (1080P).","released":"2026-04-27","apiAvailable":true,"url":"https://www.alibabacloud.com/help/en/model-studio/happyhorse-1-0-t2v","notes":"Alibaba ATH video model that topped the Artificial Analysis Video Arena in April 2026. Despite early open-source claims, no official weights have been published; access is API-only via Model Studio and partners such as fal.","maxOutput":"15s @ 1080p","features":["synchronized audio","image-to-video","reference-to-video","video edit"]},{"id":"runway-gen-4.5","name":"Runway Gen-4.5","provider":"Runway","modality":"video","pricingUnit":"per_second_video","pricingAmount":0.12,"pricingNote":"12 credits per second on the Runway API (credits are $0.01 each). Gen-4 Turbo is 5 credits ($0.05) per second. ProRes and HDR outputs add a per-second surcharge.","released":"2025-12-01","apiAvailable":true,"url":"https://runwayml.com/research/introducing-runway-gen-4.5","notes":"Runway flagship for text-to-video and image-to-video, with true HDR and ProRes delivery options. The Runway API also resells Veo 3.1, Seedance, HappyHorse, and other third-party models on the same credits.","maxOutput":"10s @ 1280x720","features":["text-to-video","image-to-video","HDR output","ProRes output"]},{"id":"luma-ray-3.2","name":"Luma Ray3.2","provider":"Luma AI","modality":"video","pricingUnit":"per_second_video","pricingAmount":0.06,"pricingNote":"Priced per clip: $0.30 per 5s at 720p ($0.06/s), $1.20 per 5s at 1080p, $0.90 per 10s at 720p. HDR is 2x and HDR plus EXR is 3x the SDR price.","released":"2026-06-09","apiAvailable":true,"url":"https://lumalabs.ai/api/pricing","notes":"First Ray model with its full control surface exposed as an API, separate from the legacy Dream Machine API that still serves Ray 2. Up to 16 keyframes per clip, native 16-bit HDR, and EXR export for VFX pipelines.","maxOutput":"10s @ 1080p (V2V up to 20s)","features":["multi-keyframe","HDR","EXR export","video-to-video"]},{"id":"pika-2.2","name":"Pika 2.2","provider":"Pika Labs","modality":"video","pricingUnit":"per_second_video","pricingAmount":0.04,"pricingNote":"$0.20 per 5s clip at 720p ($0.04/s) and $0.45 at 1080p ($0.09/s) via fal, Pika's API partner. Pika's own developer API lists Pika 2.5 at the same rates plus a $10/month membership.","released":"2025-02-04","apiAvailable":true,"url":"https://pika.art","notes":"Effects-first product (Pikaffects, Pikaswaps, Pikadditions). Pika 2.5 is the current model on the Pika API; 2.2 remains available through fal.","maxOutput":"10s @ 1080p","features":["video effects","image-to-video","lip-sync"]},{"id":"flux-3-video","name":"FLUX 3 Video","provider":"Black Forest Labs","modality":"video","pricingUnit":"per_second_video","pricingAmount":0.17,"pricingNote":"$0.17 per second at HD, $0.29 at Full HD for text or image to video; drafts $0.06/s. Video continuation $0.43 to $0.54/s. Audio included at no extra charge.","released":"2026-08-04","apiAvailable":true,"url":"https://bfl.ai/blog/flux-3-video","notes":"BFL's first video model, generally available on the BFL API after a July 23, 2026 early-access launch. Generates native audio (dialogue, effects, ambience) with the frames; Full HD comes via upscaling. FLUX 3 Image and open-weight FLUX 3 Dev are on the roadmap.","maxOutput":"20s @ 1080p with audio","features":["native audio","image-to-video","video continuation","draft mode"]},{"id":"gemini-omni-flash","name":"Gemini Omni Flash","provider":"Google","modality":"video","pricingUnit":"per_second_video","pricingAmount":0.1,"pricingNote":"Token-priced at $17.50 per 1M video output tokens (with audio), plus $1.50 per 1M input tokens. Google bills 5,792 tokens per second at 720p (about $0.10/s); 1080p (8,688 tokens/s) is about $0.15/s and 4K (17,376 tokens/s) about $0.30/s.","released":"2026-08-27","apiAvailable":true,"url":"https://ai.google.dev/gemini-api/docs/models/gemini-omni-flash","notes":"Conversational video generation and editing: create a clip, then refine it in natural language through the Interactions API. GA as gemini-omni-1.1-flash after a June 30, 2026 preview; the preview id shuts down October 22, 2026, a date Google pushed back from September 30. Google now routes the retiring Veo 3.1 preview ids here too, making this the Gemini API video path. Adds extension, first and last frame interpolation, and 1080p or 4K via upscaling. Text output is billed separately at $9 per 1M tokens.","maxOutput":"10s @ 4K (upscaled)","features":["conversational editing","extend","interpolation","native audio"]},{"id":"elevenlabs-v3","name":"Eleven v3","provider":"ElevenLabs","modality":"tts","pricingUnit":"per_1k_chars","pricingAmount":0.1,"pricingNote":"$0.10 per 1k characters on the API at every plan tier (plans bundle included characters, from $6/month Starter). Eleven v3 Conversational and Flash are $0.05.","released":"2025-06-04","apiAvailable":true,"url":"https://elevenlabs.io/pricing/api","notes":"Most expressive ElevenLabs model, now out of alpha. Supports audio tags ([whisper], [laughs]) and multi-speaker dialogue across 70+ languages. Not tuned for realtime; use Eleven v3 Conversational for live agents.","maxOutput":"70+ languages, 5,000 chars per request","features":["voice cloning","audio tags","text to dialogue","streaming"]},{"id":"elevenlabs-v3-conversational","name":"Eleven v3 Conversational","provider":"ElevenLabs","modality":"tts","pricingUnit":"per_1k_chars","pricingAmount":0.05,"pricingNote":"$0.05 per 1k characters on the API, half the price of standard Eleven v3.","released":"2026-08-19","apiAvailable":true,"url":"https://elevenlabs.io/docs/overview/models","notes":"Realtime variant of Eleven v3, generally available August 19, 2026. Keeps audio tags and 70+ language coverage at roughly 280ms latency, with a Text to Dialogue WebSocket for multi-speaker streaming.","maxOutput":"70+ languages, ~280ms latency","features":["audio tags","streaming","real-time","text to dialogue websocket"]},{"id":"cartesia-sonic-3.6","name":"Cartesia Sonic 3.6","provider":"Cartesia","modality":"tts","pricingUnit":"per_1k_chars","pricingAmount":0.05,"pricingNote":"Credit-based at about 1 credit per character. Pro plan is $5/month for 100K credits ($0.05 per 1k chars); Startup ($49 for 1.25M) and Scale ($299 for 8M) work out near $0.04.","released":"2026-08-27","apiAvailable":true,"url":"https://docs.cartesia.ai/build-with-cartesia/tts-models/latest","notes":"Cartesia's current realtime TTS model, GA with 44 languages (adds Odia and Urdu). Sonic 2, Sonic Turbo, and the first Sonic 3 snapshot sunset on October 20, 2026.","maxOutput":"44 languages","features":["instant voice cloning","streaming","websocket","dated snapshots"]},{"id":"openai-tts-hd","name":"OpenAI TTS-1-HD","provider":"OpenAI","modality":"tts","pricingUnit":"per_1k_chars","pricingAmount":0.03,"pricingNote":"$0.030 per 1k chars (TTS-1-HD), $0.015 per 1k chars (TTS-1). Both shut down January 6, 2027, and the gpt-4o-mini-tts snapshots OpenAI used to steer new work toward go the same day.","released":"2023-11-06","apiAvailable":true,"url":"https://developers.openai.com/api/docs/guides/text-to-speech","notes":"Deprecated October 1, 2026 along with tts-1, gpt-4o-mini-tts-2025-03-20 and gpt-4o-mini-tts-2025-12-15: the whole dedicated text-to-speech endpoint retires on January 6, 2027. OpenAI names gpt-realtime-2.1-mini as the replacement, which is a realtime speech model rather than a drop-in speech endpoint, so this is an API change and not a model swap. Nine fixed voices; Azure Foundry retires its tts and tts-hd copies three weeks earlier, on December 15, 2026.","maxOutput":"9 voices, 57 languages","features":["9 fixed voices","streaming","multiple formats"]},{"id":"deepgram-aura-2","name":"Deepgram Aura-2","provider":"Deepgram","modality":"tts","pricingUnit":"per_1k_chars","pricingAmount":0.03,"pricingNote":"$0.030 per 1k chars pay-as-you-go, $0.027 on the Growth plan (prepaid annual credits). The newer Flux TTS now bills at $0.045 per 1k chars pay-as-you-go and $0.0405 on Growth, with Deepgram matching every $1 of Flux TTS spend with $1 of credit up to $500 through December 31, 2026.","released":"2025-02-11","apiAvailable":true,"url":"https://deepgram.com/pricing","notes":"Low-latency TTS built for voice agents, and still half the price of Flux TTS. Pairs with Deepgram Nova-3 or Flux STT for a full voice-agent loop. New accounts get $200 in credit that does not expire until used.","maxOutput":"90 voices, 7 languages (41 English)","features":["low latency","streaming","real-time"]},{"id":"deepgram-nova-3","name":"Deepgram Nova-3","provider":"Deepgram","modality":"stt","pricingUnit":"per_minute_audio","pricingAmount":0.0043,"pricingNote":"$0.0043 per minute pre-recorded (monolingual, pay-as-you-go), $0.0036 on Growth. Streaming now lists $0.0048/min pay-as-you-go and $0.0042 on Growth, still flagged as a limited-time promotional rate with no stated end date. Multilingual is $0.0052/min pre-recorded and $0.0058/min streaming.","released":"2025-01-23","apiAvailable":true,"url":"https://deepgram.com/pricing","notes":"Deepgram flagship STT for batch and streaming, with automatic language detection and keyterm prompting. Deepgram Flux ($0.0065 to $0.0078/min streaming) adds built-in turn detection for voice agents.","maxOutput":"WER 6.84% (English), 45+ languages","features":["streaming","real-time","diarization","code-switching"]},{"id":"gpt-4o-transcribe","name":"GPT-4o Transcribe","provider":"OpenAI","modality":"stt","pricingUnit":"per_minute_audio","pricingAmount":0.006,"pricingNote":"$0.006 per minute ($2.50 per 1M audio input tokens). gpt-4o-mini-transcribe is $0.003/min. Both shut down February 26, 2027.","released":"2025-03-20","apiAvailable":true,"url":"https://platform.openai.com/docs/guides/speech-to-text","notes":"Deprecated August 26, 2026. OpenAI now recommends gpt-transcribe ($0.0045/min) for files and gpt-live-transcribe ($0.017/min) for live audio. Azure Foundry retired its gpt-4o-transcribe and gpt-4o-mini-transcribe 2025-03-20 snapshots on October 15, 2026, four months ahead of the OpenAI API.","maxOutput":"WER 6.7% (English), 100+ languages","features":["streaming","low hallucination","multilingual"]},{"id":"gpt-transcribe","name":"GPT Transcribe","provider":"OpenAI","modality":"stt","pricingUnit":"per_minute_audio","pricingAmount":0.0045,"pricingNote":"$0.0045 per minute of audio. Companion gpt-live-transcribe for low-latency live streams is $0.017/min.","released":"2026-07-28","apiAvailable":true,"url":"https://developers.openai.com/api/docs/models/gpt-transcribe","notes":"OpenAI's recommended file transcription model, replacing whisper-1 and the gpt-4o transcribe family (all shut down February 26, 2027). Accepts free-form context, keyword hints, and multiple expected languages, and returns detected languages.","maxOutput":"25 MB files, language detection","features":["keyword hints","multilingual","streaming","realtime turns"]},{"id":"whisper-large-v3","name":"Whisper Large v3","provider":"OpenAI","modality":"stt","pricingUnit":"per_minute_audio","pricingAmount":0,"pricingNote":"Open weights, free to self-host. Groq hosts it at $0.111/hour (about $0.0019/min) and whisper-large-v3-turbo at $0.04/hour. OpenAI's hosted whisper-1 shuts down February 26, 2027.","released":"2023-11-06","apiAvailable":true,"url":"https://github.com/openai/whisper","notes":"Open-weights workhorse under the MIT license. Groq hosting of the Turbo variant at about $0.0007/min is among the cheapest hosted STT. Deepgram also serves Whisper Large at $0.0048/min.","maxOutput":"WER 7.4% (English), 99 languages","features":["open weights","mit license","multilingual"]},{"id":"assemblyai-universal-2","name":"AssemblyAI Universal-2","provider":"AssemblyAI","modality":"stt","pricingUnit":"per_minute_audio","pricingAmount":0.0025,"pricingNote":"$0.15 per hour (about $0.0025/min) pre-recorded. Universal-Streaming is also $0.15/hr but billed on WebSocket session time. Add-ons such as diarization, entity detection, and sentiment cost extra.","released":"2025-01-15","apiAvailable":true,"url":"https://www.assemblyai.com","notes":"Budget tier behind Universal-3.5 Pro ($0.21/hr pre-recorded, $0.45/hr realtime). Built-in summarization and auto chapters are deprecated in favor of the AssemblyAI LLM Gateway.","maxOutput":"WER 5.6% (English), 99 languages","features":["sentiment","entity detection","topics","streaming"]},{"id":"google-chirp-2","name":"Chirp 2","provider":"Google","modality":"stt","pricingUnit":"per_minute_audio","pricingAmount":0.016,"pricingNote":"$0.016 per minute on Speech-to-Text V2 for the first 500K minutes a month, tiering down to $0.004 above 2M. Dynamic batch recognition is $0.003/min.","released":"2024-11-08","apiAvailable":true,"url":"https://docs.cloud.google.com/speech-to-text/docs/models/chirp-2","notes":"Google Cloud Speech-to-Text V2 model with broad multilingual coverage and speech translation. Chirp 3 is now the latest Chirp generation, adding diarization and automatic language detection.","maxOutput":"125+ languages","features":["multilingual","streaming","speech translation","enterprise SLA"]},{"id":"gemini-3.5-transcribe","name":"Gemini 3.5 Transcribe","provider":"Google","modality":"stt","pricingUnit":"per_minute_audio","pricingAmount":0.005,"pricingNote":"Token-priced ($2 per 1M audio input tokens, $12 per 1M text output tokens); Google estimates about $0.005 per minute blended. The streaming Live variant is about $0.009/min.","released":"2026-08-26","apiAvailable":true,"url":"https://ai.google.dev/gemini-api/docs/models/gemini-3.5-transcribe","notes":"Dedicated Gemini speech-to-text model (GA) with speaker diarization, word-level timestamps, and custom vocabulary biasing up to 1,000 terms. A separate gemini-3.5-transcribe-live model streams over WebSockets.","maxOutput":"85+ languages, 1 hour per request","features":["diarization","word timestamps","custom vocabulary","code-switching"]},{"id":"gpt-realtime-2.1","name":"GPT-Realtime-2.1","provider":"OpenAI","modality":"stt","pricingUnit":"per_minute_audio","pricingAmount":0.0192,"pricingNote":"Speech-to-speech model listed under stt. Token-priced at $32 per 1M audio input and $64 per 1M audio output tokens; at OpenAI's 1 token per 100ms of user audio and 1 per 50ms of model audio, that is $0.0192 per input minute and $0.0768 per output minute, before context re-billing on later turns.","released":"2026-07-06","apiAvailable":true,"url":"https://developers.openai.com/api/docs/models/gpt-realtime-2.1","notes":"Realtime voice-agent model with configurable reasoning and tool use, updating the May 2026 GPT-Realtime-2 with better alphanumeric recognition, noise handling, and interruptions. gpt-realtime-2.1-mini is $10/$20 per 1M audio tokens. OpenAI also sells GPT-Live 1 full-duplex voice sessions at $0.05/min plus backend usage.","maxOutput":"128K context, 32K output tokens","features":["speech-to-speech","reasoning","function calling","image input"]}]}