# Cloud providers compared. Dimmy Help. > Cloud providers compared. Step-by-step help for Dimmy, the native voice dictation app for macOS, Windows, and Linux. Source: https://dimmy.app/help/cloud-providers/ [← Back to Help center](https://dimmy.app/help/) Providers & Models # Cloud providers compared Groq is fastest, OpenAI is the gold standard, Deepgram leads on streaming. All optional. Bring your own key. Dimmy ships a catalogue of 8 cloud providers, plus a Custom endpoint for any OpenAI-compatible API. You bring the key, Dimmy talks to the API directly. We never proxy your traffic. | Provider | Does | Free tier | Best for | | ---------- | ---------------- | --------------------- | ------------------------------------------------------------------------------------------------------------------------------- | | Groq | Speech + rewrite | Yes, rate-limited | The fastest speech to text, and the shipped cloud default (Whisper Large v3 Turbo). Rewrite: GPT-OSS 120B and 20B, Qwen3.6 27B. | | OpenAI | Speech + rewrite | Pay-per-use | Best on noisy audio, now with gpt-transcribe rather than whisper-1\. Rewrite: the GPT-5.6 line, Sol at the top. | | Anthropic | Rewrite only | Pay-per-use | No speech to text at all. Claude Opus 5 at the top, Sonnet 5 balanced, Haiku 5.5 for speed. | | Gemini | Speech + rewrite | Yes | Long context. Gemini 3.1 Pro at the top, 3.6 and 3.5 Flash for speed. | | Deepgram | Speech only | Free credits to start | Real-time streaming and the lowest end-to-end latency, on nova-3. | | Together | Speech + rewrite | Pay-per-use | Qwen3.7 Max, Kimi K3, GLM-5.2, DeepSeek V4 Pro. Speech: Parakeet TDT v3 and Whisper Large v3. | | Fireworks | Speech + rewrite | Pay-per-use | Kimi K3, GLM-5.2, DeepSeek V4 Pro, MiniMax M3, Qwen3.7 Plus. Speech: Whisper v3 Turbo. | | OpenRouter | Rewrite only | Two free models | Llama 3.3 70B and DeepSeek R1, both on free ids. Cannot be used for meeting recaps. | | Custom | Speech + rewrite | Yours | Any OpenAI-compatible HTTPS endpoint. Self-hosted, or a provider we do not list. | ## Not every provider does everything * Six of the eight do speech to text. Anthropic and OpenRouter are rewrite-only, they have no transcription endpoint. * Deepgram is the mirror image: speech to text only, no rewrite and no recap. * OpenRouter's two models are not offered for meeting recaps. Everything else in the catalogue is. * Models are kept in one catalogue compiled into the app, so Windows and macOS read the same list and cannot drift apart. Retired model ids are removed rather than left to fail at call time. ## Streaming dictation Two cloud backends can transcribe while you are still talking: OpenAI Realtime, which Dimmy prefers, and Deepgram nova-3\. Nobody else in the list streams. With neither key present, streaming falls back to the local chunked transcriber. Which one is used follows the speech provider you already configured, there is no second setting to keep in sync. ## File size limits Long recordings are split before upload when the provider will not take them whole. The ceilings differ a lot, and Gemini's is the one you will notice. | Provider | Largest upload | | -------------------- | -------------- | | Deepgram | 2 GB | | Fireworks, Together | 100 MB | | Groq, OpenAI, Custom | 25 MB | | Gemini | 20 MB | | Local | No limit | warning A Dimmy plan does not unlock any model. There is no Dimmy-hosted proxy: every cloud model here runs on your own API key or your own provider plan, on any plan, including none. See [Which model should you use?](https://dimmy.app/help/choose-your-model/). ## Adding a provider Open Settings → Providers & keys and find the provider card. Add your API key in the key field. The key is stored encrypted on your device. See [Add your API key](https://dimmy.app/help/api-keys/). The On-device card on the same page lists your local models with a green check next to the ones already downloaded, see [Choose a Whisper model](https://dimmy.app/help/whisper-models/). ## Mixing local and cloud You can run speech to text locally and the rewrite in the cloud, or the reverse, and the meeting recap is a third independent choice with its own provider and its own key. A common setup is local Whisper for transcription, which never leaves the machine, and a cloud flagship for the recap. See [the on-device LLM](https://dimmy.app/help/local-llm/) and [Meeting recaps](https://dimmy.app/help/use-meeting-recaps/). info All cloud calls are logged to your Dimmy log file with the provider name (never the key, never the payload). Check the log if you want to audit. Was this helpful? ## In this category [ Providers & ModelsWhich model should you use?Four questions decide it: what machine you have, what must not leave it, which subscriptions you already pay for, and which language you speak. Answer them and the rest follows.](https://dimmy.app/help/choose-your-model/) [ Providers & ModelsLocal mode: transcription on your own machineFour speech engines that run on your device. No API key, no upload, no per-minute cost. Works on a plane.](https://dimmy.app/help/local-mode/) [ Providers & ModelsAdd your API keyOpen Settings → Providers & keys, find the provider card, paste the key. It is encrypted on your machine. You can update keys at any time.](https://dimmy.app/help/api-keys/) [ Providers & ModelsChoose a local modelThe whole on-device catalogue: nine Whisper sizes, Parakeet TDT v3, four Qwen3-ASR variants and Whistle. What each one weighs and what it runs on.](https://dimmy.app/help/whisper-models/) [ Providers & ModelsOptional on-device LLMRun the LLM enhancement pass locally via llama.cpp. No cloud, no key, no per-token cost. Pick a model to match your GPU memory.](https://dimmy.app/help/local-llm/) Last updated: 2026-05-22 On this page 1. [01 Not every provider does everything](#not-every-provider-does-everything) 2. [02 Streaming dictation](#streaming-dictation) 3. [03 File size limits](#file-size-limits) 4. [04 Adding a provider](#adding-a-provider) 5. [05 Mixing local and cloud](#mixing-local-and-cloud)