# Models (/models)

Start with the speech model ScribeKey suggests during setup. It uses your phone's memory to help you choose. These models run on your phone, and each works offline once it is downloaded.

Cloud speech sends audio to a provider you set up. Cloud skills send text to a provider only when you choose them. [How models reach your phone](/models/how-models-reach-your-phone) explains where the list comes from.

## English [#english]

Smallest download first. Setup suggests one for your phone's memory.

### Moonshine Tiny [#model-moonshine-v2-tiny-en]

**Small English download.** An English model with a small download and memory requirement.

- **Languages:** English
- **Text:** Arrives when you stop
- **Download:** 43 MB
- **Phone memory:** 2 GB or more
- Runs on Android 9 or later
- **Size:** 26M parameters
- **Licence:** [MIT](https://huggingface.co/csukuangfj2/sherpa-onnx-moonshine-tiny-en-quantized-2026-02-27/blob/d1e6c30921780b8508d04b492dfb3ce8a51605d4/LICENSE)
- **Model files:** [Hugging Face](https://huggingface.co/csukuangfj2/sherpa-onnx-moonshine-tiny-en-quantized-2026-02-27/tree/d1e6c30921780b8508d04b492dfb3ce8a51605d4)
- **Catalogue id:** `moonshine-v2-tiny-en`

### Moonshine Base [#model-moonshine-v2-base-en]

**English on phones with less memory.** A larger Moonshine for English that still needs little memory. Text appears when you finish speaking.

- **Languages:** English
- **Text:** Arrives when you stop
- **Download:** 135 MB
- **Phone memory:** 2 GB or more
- Runs on Android 9 or later
- **Size:** 58M parameters
- **Licence:** [MIT](https://huggingface.co/csukuangfj2/sherpa-onnx-moonshine-base-en-quantized-2026-02-27/blob/8f4d6c58c03d40bcea40043bb7120a878f2bbef6/LICENSE)
- **Model files:** [Hugging Face](https://huggingface.co/csukuangfj2/sherpa-onnx-moonshine-base-en-quantized-2026-02-27/tree/8f4d6c58c03d40bcea40043bb7120a878f2bbef6)
- **Catalogue id:** `moonshine-v2-base-en`

### Distil-Whisper Small [#model-distil-small-en]

**English from a smaller Whisper model.** A distilled Whisper model for English. Text appears when you finish speaking, with punctuation and capitals.

- **Languages:** English
- **Text:** Arrives when you stop
- **Download:** 286 MB
- **Phone memory:** 3 GB or more
- Runs on Android 9 or later
- **Size:** 190M parameters
- **Licence:** [MIT](https://huggingface.co/distil-whisper/distil-small.en)
- **Model files:** [Hugging Face](https://huggingface.co/csukuangfj/sherpa-onnx-whisper-distil-small.en/tree/0492324bca9e12a6fca0089bb846f2dd723b50d8)
- **Catalogue id:** `distil-small-en`

### Parakeet Compact [#model-parakeet-110m]

**English on mid-range phones.** Transcribes English when you finish speaking, with punctuation and capitals.

- **Languages:** English
- **Text:** Arrives when you stop
- **Download:** 437 MB
- **Phone memory:** 3 GB or more
- Runs on Android 9 or later
- **Size:** 110M parameters
- **Licence:** [CC-BY-4.0](https://huggingface.co/nvidia/parakeet-tdt_ctc-110m)
- **Model files:** [Hugging Face](https://huggingface.co/csukuangfj/sherpa-onnx-nemo-parakeet_tdt_ctc_110m-en-36000/tree/3af92f152d32c836acabf38f4c993bc96b80eb2d)
- **Catalogue id:** `parakeet-110m`

### Parakeet English [#model-parakeet-unified-0.6b]

**English dictation with punctuation.** Transcribes English with punctuation and timestamps.

- **Languages:** English
- **Text:** Arrives when you stop
- **Download:** 633 MB
- **Phone memory:** 6 GB or more
- Runs on Android 12 or later
- **Size:** 600M parameters
- **Licence:** [NVIDIA Open Model License](https://www.nvidia.com/en-us/agreements/enterprise-software/nvidia-open-model-license/)
- **Model files:** [Hugging Face](https://huggingface.co/csukuangfj2/sherpa-onnx-nemo-parakeet-unified-en-0.6b-int8-non-streaming/tree/8c3a10fb13408c7a7054f6898958bf1c64a8d6c7)
- **Catalogue id:** `parakeet-unified-0.6b`

## Live words [#live-words]

Words appear while you speak. The text is finalised when you stop.

### Zipformer English Live [#model-zipformer-en-live]

**Live English on 3 GB phones.** Shows English words while you speak. Punctuation is added when you stop.

- **Languages:** English
- **Text:** Appears as you speak
- **Download:** 188 MB
- **Phone memory:** 3 GB or more
- Runs on Android 9 or later
- **Size:** 70M parameters
- **Licence:** [Apache-2.0](https://huggingface.co/csukuangfj/sherpa-onnx-streaming-zipformer-en-2023-06-21)
- **Model files:** [Hugging Face](https://huggingface.co/csukuangfj/sherpa-onnx-streaming-zipformer-en-2023-06-21/tree/9a65b6ea94c311ca770c2bf895b30f456a22d703)
- **Catalogue id:** `zipformer-en-live`

### Nemotron English Live [#model-nemotron-en-0.6b]

**English words appear as you speak.** Shows English words while you speak, with punctuation.

- **Languages:** English
- **Text:** Appears as you speak
- **Download:** 632 MB
- **Phone memory:** 6 GB or more
- Runs on Android 12 or later
- **Size:** 600M parameters
- **Licence:** [NVIDIA Open Model License](https://www.nvidia.com/en-us/agreements/enterprise-software/nvidia-open-model-license/)
- **Model files:** [Hugging Face](https://huggingface.co/csukuangfj2/sherpa-onnx-nemotron-speech-streaming-en-0.6b-560ms-int8-2026-04-25/tree/52056fdc070914a48dcd68b31b44d6a6f5b85902)
- **Catalogue id:** `nemotron-en-0.6b`

### Nemotron 3.5 Multilingual Live [#model-nemotron-3.5-0.6b-560ms]

**Live dictation in multiple languages.** Shows words while you speak in multiple locales and picks the language for you.

- **Languages:** Multiple languages
  - Ready: English, Spanish, French, Italian, Portuguese, Dutch, German, Turkish, Russian, Arabic, Hindi, Japanese, Korean, Vietnamese and Ukrainian
  - Broad coverage: Polish, Swedish, Czech, Norwegian Bokmål, Danish, Bulgarian, Finnish, Croatian, Slovak, Chinese, Hungarian, Romanian and Estonian
  - Weaker, still improving: Greek, Lithuanian, Latvian, Maltese, Slovenian, Hebrew, Thai and Norwegian Nynorsk
- **Text:** Appears as you speak
- **Download:** 651 MB
- **Phone memory:** 6 GB or more
- Runs on Android 12 or later
- **Size:** 600M parameters
- **Licence:** [OpenMDW-1.1](https://openmdw.ai/license/1-1/)
- **Model files:** [Hugging Face](https://huggingface.co/csukuangfj2/sherpa-onnx-nemotron-3.5-asr-streaming-0.6b-560ms-int8-2026-06-11/tree/ab43d895f5985b1bbab8b6eac8607fcdc05343f3)
- **Catalogue id:** `nemotron-3.5-0.6b-560ms`

## Many languages [#many-languages]

A model with language detection recognises the language you're speaking.

### Canary Multilingual [#model-canary-180m]

**Four European languages in a small download.** Transcribes European languages with punctuation and language detection.

- **Languages:** English, Spanish, German and French
- **Text:** Arrives when you stop
- **Download:** 198 MB
- **Phone memory:** 4 GB or more
- Runs on Android 9 or later
- **Size:** 180M parameters
- **Licence:** [CC-BY-4.0](https://huggingface.co/nvidia/canary-180m-flash)
- **Model files:** [Hugging Face](https://huggingface.co/csukuangfj/sherpa-onnx-nemo-canary-180m-flash-en-es-de-fr-int8/tree/9077164e0d3dd1d5353743e89ceaa1d3a770838c)
- **Catalogue id:** `canary-180m`

### SenseVoice Small [#model-sensevoice-small]

**Chinese, Cantonese, Japanese and Korean.** Transcribes Chinese, Cantonese, English, Japanese and Korean with punctuation, and detects which one you speak.

- **Languages:** Chinese, Cantonese, English, Japanese and Korean
- **Detects the language:** Yes
- **Text:** Arrives when you stop
- **Download:** 229 MB
- **Phone memory:** 3 GB or more
- Runs on Android 9 or later
- **Size:** 234M parameters
- **Licence:** [Apache-2.0](https://huggingface.co/FunAudioLLM/SenseVoiceSmall)
- **Model files:** [Hugging Face](https://huggingface.co/csukuangfj/sherpa-onnx-sense-voice-zh-en-ja-ko-yue-2024-07-17/tree/2365baeacb507f821a0c8120fcee3d484dba7a07)
- **Catalogue id:** `sensevoice-small`

### Parakeet Ultra [#model-parakeet-ultra-0.6b]

**Further-trained Parakeet for European languages.** Transcribes European languages.

- **Languages:** 25 European languages
  - English, Spanish, French, German, Bulgarian, Croatian, Czech, Danish, Dutch, Estonian, Finnish, Greek, Hungarian, Italian, Latvian, Lithuanian, Maltese, Polish, Portuguese, Romanian, Slovak, Slovenian, Swedish, Russian and Ukrainian
- **Detects the language:** Yes
- **Text:** Arrives when you stop
- **Download:** 601 MB
- **Phone memory:** 6 GB or more
- Runs on Android 12 or later
- **Size:** 600M parameters
- **Licence:** [CC-BY-4.0](https://huggingface.co/moondream/parakeet-ultra)
- **Model files:** [Hugging Face](https://huggingface.co/mldecode/parakeet-ultra-onnx-int8/tree/3282a6e32885b431c1543d58c7710e6e3412eac0)
- **Catalogue id:** `parakeet-ultra-0.6b`

### Parakeet Multilingual [#model-parakeet-0.6b-v3]

**European-language dictation.** Transcribes European languages with punctuation, timestamps and language detection.

- **Languages:** 25 European languages
  - English, Spanish, French, German, Bulgarian, Croatian, Czech, Danish, Dutch, Estonian, Finnish, Greek, Hungarian, Italian, Latvian, Lithuanian, Maltese, Polish, Portuguese, Romanian, Slovak, Slovenian, Swedish, Russian and Ukrainian
- **Detects the language:** Yes
- **Text:** Arrives when you stop
- **Download:** 640 MB
- **Phone memory:** 6 GB or more
- Runs on Android 12 or later
- **Size:** 600M parameters
- **Licence:** [CC-BY-4.0](https://huggingface.co/nvidia/parakeet-tdt-0.6b-v3)
- **Model files:** [Hugging Face](https://huggingface.co/csukuangfj/sherpa-onnx-nemo-parakeet-tdt-0.6b-v3-int8/tree/2bda32ec70b097a55adaa07d9a7173915b43cc78)
- **Catalogue id:** `parakeet-0.6b-v3`

## Experimental [#experimental]

The catalogue marks these models Experimental.

### Omnilingual 300M [#model-omnilingual-asr-300m]

**Languages other models do not cover.** A broad-coverage model without punctuation.

- **Status:** Experimental
- **Languages:** Over 1,600 languages
- **Text:** Arrives when you stop
- **Download:** 349 MB
- **Phone memory:** 4 GB or more
- Runs on Android 9 or later
- **Size:** 300M parameters
- **Licence:** [Apache-2.0](https://huggingface.co/facebook/omniASR-CTC-300M)
- **Model files:** [Hugging Face](https://huggingface.co/csukuangfj2/sherpa-onnx-omnilingual-asr-1600-languages-300M-ctc-int8-2025-11-12/tree/6fc542a3b0661c8278cca1230c34deb989f31202)
- **Catalogue id:** `omnilingual-asr-300m`

### Parakeet Redux [#model-parakeet-redux-0.6b]

**Smaller download for 25 languages.** A 1.58-bit Parakeet for 25 European languages in a smaller download, with punctuation, timestamps and language detection.

- **Status:** Experimental
- **Languages:** Multiple languages
  - English, Spanish, French, German, Bulgarian, Croatian, Czech, Danish, Dutch, Estonian, Finnish, Greek, Hungarian, Italian, Latvian, Lithuanian, Maltese, Polish, Portuguese, Romanian, Slovak, Slovenian, Swedish, Russian and Ukrainian
- **Detects the language:** Yes
- **Text:** Arrives when you stop
- **Download:** 398 MB
- **Phone memory:** 6 GB or more
- Runs on Android 12 or later
- **Size:** 600M parameters
- **Licence:** [CC-BY-4.0](https://huggingface.co/moondream/parakeet-redux)
- **Model files:** [Hugging Face](https://huggingface.co/cohearo/sherpa-onnx-nemo-parakeet-redux/tree/5168f843ba8f01bf24594f58e5d21f28f63fbb44)
- **Catalogue id:** `parakeet-redux-0.6b`

### Qwen3 ASR [#model-qwen3-asr-0.6b]

**Chinese dialects and multiple languages.** Transcribes multiple languages and Chinese dialects.

- **Status:** Experimental
- **Languages:** 30 languages and Chinese dialects
- **Text:** Arrives when you stop
- **Download:** 942 MB
- **Phone memory:** 6 GB or more
- Runs on Android 9 or later
- **Size:** 600M parameters
- **Licence:** [Apache-2.0](https://huggingface.co/Qwen/Qwen3-ASR-0.6B)
- **Model files:** [Hugging Face](https://huggingface.co/csukuangfj2/sherpa-onnx-qwen3-asr-0.6B-int8-2026-03-25/tree/68818b2313fe77bd06f6a7c5068ff3ef59d02b8a)
- **Catalogue id:** `qwen3-asr-0.6b`

## Older models [#older-models]

The catalogue marks these models Deprecated or Retired.

### Moonshine Tiny v1 [#model-moonshine-tiny]

**Existing installs only.** The original Moonshine Tiny, kept so existing installs keep working.

- **Status:** Deprecated
- **Languages:** English
- **Text:** Arrives when you stop
- **Download:** 119 MB
- **Phone memory:** 2 GB or more
- Runs on Android 9 or later
- **Size:** 26M parameters
- **Licence:** [MIT](https://huggingface.co/moonshine-ai/moonshine-tiny)
- **Model files:** [Hugging Face](https://huggingface.co/csukuangfj/sherpa-onnx-moonshine-tiny-en-int8/tree/bf2b762c076d8ea61e2af0b3851c9564fb77552e)
- **Catalogue id:** `moonshine-tiny`

### Moonshine Base v1 [#model-moonshine-base]

**Existing installs only.** The original Moonshine Base, kept so existing installs keep working.

- **Status:** Deprecated
- **Languages:** English
- **Text:** Arrives when you stop
- **Download:** 274 MB
- **Phone memory:** 2 GB or more
- Runs on Android 9 or later
- **Size:** 58M parameters
- **Licence:** [MIT](https://huggingface.co/moonshine-ai/moonshine-base)
- **Model files:** [Hugging Face](https://huggingface.co/csukuangfj/sherpa-onnx-moonshine-base-en-int8/tree/052b0798ad1bf046a140fdd4efcd9426530fa3f5)
- **Catalogue id:** `moonshine-base`

### Parakeet English v2 [#model-parakeet-0.6b-v2]

**Existing installs only.** An earlier Parakeet English model, kept so existing installs keep working.

- **Status:** Retired
- **Languages:** English
- **Text:** Arrives when you stop
- **Download:** 631 MB
- **Phone memory:** 6 GB or more
- Runs on Android 12 or later
- **Size:** 600M parameters
- **Licence:** [CC-BY-4.0](https://huggingface.co/nvidia/parakeet-tdt-0.6b-v2)
- **Model files:** [Hugging Face](https://huggingface.co/csukuangfj/sherpa-onnx-nemo-parakeet-tdt-0.6b-v2-int8/tree/1ab9323565ddb038682214b292f588070a538ce2)
- **Catalogue id:** `parakeet-0.6b-v2`

### Parakeet English Live preview [#model-parakeet-unified-0.6b-streaming-560ms]

**Existing installs only.** An early live English preview, kept so existing installs keep working.

- **Status:** Retired
- **Languages:** English
- **Text:** Appears as you speak
- **Download:** 633 MB
- **Phone memory:** 6 GB or more
- Runs on Android 12 or later
- **Size:** 600M parameters
- **Licence:** [NVIDIA Open Model License](https://www.nvidia.com/en-us/agreements/enterprise-software/nvidia-open-model-license/)
- **Model files:** [Hugging Face](https://huggingface.co/csukuangfj2/sherpa-onnx-nemo-parakeet-unified-en-0.6b-int8-streaming-560ms/tree/7551fd26fc810cc1e4e043e608db4d13b59be31e)
- **Catalogue id:** `parakeet-unified-0.6b-streaming-560ms`

## By language [#by-language]

Find a model for your language, then try a short note. Accuracy varies by language and accent.

| Language | Models |
| --- | --- |
| Arabic | Nemotron 3.5 Multilingual Live |
| Bulgarian | Parakeet Multilingual, Parakeet Ultra, Nemotron 3.5 Multilingual Live |
| Cantonese | SenseVoice Small |
| Chinese | SenseVoice Small, Nemotron 3.5 Multilingual Live |
| Croatian | Parakeet Multilingual, Parakeet Ultra, Nemotron 3.5 Multilingual Live |
| Czech | Parakeet Multilingual, Parakeet Ultra, Nemotron 3.5 Multilingual Live |
| Danish | Parakeet Multilingual, Parakeet Ultra, Nemotron 3.5 Multilingual Live |
| Dutch | Parakeet Multilingual, Parakeet Ultra, Nemotron 3.5 Multilingual Live |
| English | Moonshine Tiny, Moonshine Base, Distil-Whisper Small, Parakeet Compact, Canary Multilingual, SenseVoice Small, Parakeet Multilingual, Parakeet Ultra, Parakeet English, Nemotron English Live, Nemotron 3.5 Multilingual Live, Zipformer English Live |
| Estonian | Parakeet Multilingual, Parakeet Ultra, Nemotron 3.5 Multilingual Live |
| Finnish | Parakeet Multilingual, Parakeet Ultra, Nemotron 3.5 Multilingual Live |
| French | Canary Multilingual, Parakeet Multilingual, Parakeet Ultra, Nemotron 3.5 Multilingual Live |
| German | Canary Multilingual, Parakeet Multilingual, Parakeet Ultra, Nemotron 3.5 Multilingual Live |
| Greek | Parakeet Multilingual, Parakeet Ultra |
| Hindi | Nemotron 3.5 Multilingual Live |
| Hungarian | Parakeet Multilingual, Parakeet Ultra, Nemotron 3.5 Multilingual Live |
| Italian | Parakeet Multilingual, Parakeet Ultra, Nemotron 3.5 Multilingual Live |
| Japanese | SenseVoice Small, Nemotron 3.5 Multilingual Live |
| Korean | SenseVoice Small, Nemotron 3.5 Multilingual Live |
| Latvian | Parakeet Multilingual, Parakeet Ultra |
| Lithuanian | Parakeet Multilingual, Parakeet Ultra |
| Maltese | Parakeet Multilingual, Parakeet Ultra |
| Norwegian Bokmål | Nemotron 3.5 Multilingual Live |
| Polish | Parakeet Multilingual, Parakeet Ultra, Nemotron 3.5 Multilingual Live |
| Portuguese | Parakeet Multilingual, Parakeet Ultra, Nemotron 3.5 Multilingual Live |
| Romanian | Parakeet Multilingual, Parakeet Ultra, Nemotron 3.5 Multilingual Live |
| Russian | Parakeet Multilingual, Parakeet Ultra, Nemotron 3.5 Multilingual Live |
| Slovak | Parakeet Multilingual, Parakeet Ultra, Nemotron 3.5 Multilingual Live |
| Slovenian | Parakeet Multilingual, Parakeet Ultra |
| Spanish | Canary Multilingual, Parakeet Multilingual, Parakeet Ultra, Nemotron 3.5 Multilingual Live |
| Swedish | Parakeet Multilingual, Parakeet Ultra, Nemotron 3.5 Multilingual Live |
| Turkish | Nemotron 3.5 Multilingual Live |
| Ukrainian | Parakeet Multilingual, Parakeet Ultra, Nemotron 3.5 Multilingual Live |
| Vietnamese | Nemotron 3.5 Multilingual Live |

## Cleanup [#cleanup]

Basic cleanup removes "um", adds capitals and full stops, and formats times and numbers. It runs on your phone and needs no model download.

Refine uses ScribeKey Fast on your phone to remove repeats and false starts, apply spoken corrections, follow your writing style and turn three or more items into bullets. It needs a 64-bit phone with at least 6 GB of memory. Other phones keep Basic cleanup.

### ScribeKey Fast [#model-scribekey-cleanup-fast]

**Fast cleanup with writing styles.** Cleans English dictation, follows writing styles and applies spoken corrections. Runs on your phone.

- **Languages:** English
- **Version:** 1.1
- **Download:** 379 MB
- **Phone memory:** 6 GB or more
- **Phone:** 64-bit
- Runs on Android 9 or later
- **Size:** 350M parameters
- **Licence:** [ScribeKey Fast + LFM v1.0](https://huggingface.co/scribekey/scribekey-cleanup-fast-GGUF/blob/4b3dce49913c15f6341393728b592a220c370efb/ADDITIONAL_TERMS.md)
- **Model card:** [Hugging Face](https://huggingface.co/scribekey/scribekey-cleanup-fast-GGUF/tree/4b3dce49913c15f6341393728b592a220c370efb)
- **Catalogue id:** `scribekey-cleanup-fast`

ScribeKey Fast is based on LFM2.5-350M by Liquid AI under the LFM Open License v1.0, and Sotto by juanquivilla under MIT.

### ScribeKey Fast cleanup examples [#scribekey-fast-cleanup-examples]

Compare the same words with Basic and Refine cleanup. These are recorded outputs. Results vary by phone and sentence.

**Example 1 of 4**

- You say: um so i was thinking we could uh meet at the cafe at three no actually four
- Basic types: So I was thinking we could meet at the cafe at three no actually four.
- ScribeKey Fast typed: So I was thinking we could meet at the cafe at four.

**Example 2 of 4**

- You say: so basically like what i'm trying to say is the client loved it and they want to sign by friday
- Basic types: So basically like what I'm trying to say is the client loved it and they want to sign by friday.
- ScribeKey Fast typed: The client loved it and they want to sign by Friday.

**Example 3 of 4**

- You say: hey sam the invoice is forty five dollars and it's due on the twenty third of november
- Basic types: Hey sam the invoice is $45 and it's due on 23 November.
- ScribeKey Fast typed: Hey Sam, the invoice is $45 and it's due on the 23rd of November.

**Example 4 of 4**

- You say: can you can you send me the slides from the meeting before lunch tomorrow
- Basic types: Can you send me the slides from the meeting before lunch tomorrow.
- ScribeKey Fast typed: Can you send me the slides from the meeting before lunch tomorrow?

## Cloud services [#cloud-services]

To use Cloud speech or Cloud skills, add your provider's API key in Settings.

- Cloud speech sends audio to your provider. Choose an OpenAI, Groq, Mistral or Microsoft Foundry preset, or enter an OpenAI-compatible URL.
- Cloud skills send text only when you choose them. Presets include OpenAI, Anthropic, Gemini, Groq and Mistral, or use any OpenAI-compatible service.

Evaluation snapshot: 28 October 2026. Each table lists the model that evaluation recommends for each provider.

### Cloud speech [#cloud-speech]

| Provider | Recommended model |
| --- | --- |
| OpenAI | gpt-transcribe |

Evaluated alternatives:

| Provider | Quality option | Speed option | Budget option |
| --- | --- | --- | --- |
| OpenAI | gpt-transcribe | gpt-transcribe | gpt-transcribe |

Scored on 156 test cases. Evaluation 7e9d79ae0663426b. Updated 9 October 2026. Source: [scribekey-app/scribekey-models/generated/cloud\_model\_tiers.json](https://github.com/scribekey-app/scribekey-models/blob/81ef5728541f052a5b1840d1172213d4b3969225/generated/cloud_model_tiers.json) at `81ef572`.

### Cloud skills [#cloud-skills]

| Provider | Recommended model |
| --- | --- |
| Anthropic | claude-haiku-5-5 |
| Cerebras | gpt-oss-120b |
| Gemini | gemini-3.1-flash-lite |
| Groq | openai/gpt-oss-120b |
| OpenAI | gpt-5.6-luna |
| OpenRouter | google/gemma-4-31b-it |

Evaluated alternatives:

| Provider | Quality option | Speed option | Budget option |
| --- | --- | --- | --- |
| Anthropic | claude-haiku-5-5 | claude-haiku-5-5 | claude-haiku-5-5 |
| Cerebras | gpt-oss-120b | gpt-oss-120b | gpt-oss-120b |
| Gemini | gemini-3.8-flash | gemini-3.1-flash-lite | gemini-3.1-flash-lite |
| Groq | openai/gpt-oss-120b | openai/gpt-oss-120b | openai/gpt-oss-20b |
| OpenAI | gpt-5.6-luna | gpt-5.4-mini | gpt-6-luna |
| OpenRouter | google/gemini-3.8-flash | openai/gpt-5.4-mini | google/gemma-4-31b-it |

Scored on 39 test cases. Evaluation d146114c42fb6bdd. Updated 9 October 2026. Source: [scribekey-app/scribekey-models/generated/cloud\_model\_tiers.json](https://github.com/scribekey-app/scribekey-models/blob/81ef5728541f052a5b1840d1172213d4b3969225/generated/cloud_model_tiers.json) at `81ef572`.