Unfragile

Fun-ASR-Nano-2512-gguf is not among the candidate products, and none of the sources mention it. The sources show general routes for running GGUF speech models: ggml, the tensor library behind GGUF, and llama.cpp, which is built on it 2; and LocalAI, an open-source engine that runs voice models locally and transcribes audio 48. Hosted speech-to-text is also available from ElevenLabs 37.

1
ggmlggml.aiverified and signed
ggml is a tensor library for machine learning that runs large models on commodity hardware, and the GGUF format is built on it 26.
Website
for agents: gguf · llama.cpp · ggml
2
LocalAIlocalai.ioverified and signed
LocalAI runs voice, LLM and other models on your own hardware and exposes a transcribe_audio action that turns speech in an audio file into text 48.
Website Docs
+6 community tools built for it
for agents: transcribe_audio · text_to_speech · list_models
3
ElevenLabselevenlabs.iofound in code, not yet checked
ElevenLabs offers speech-to-text through its speech_to_text action, which transcribes audio to text using Scribe v2 37.
Website Docs Pricing Get started
official:MCP server · ElevenLabs+58 community tools built for it
for agents: speech_to_text · speech_to_speech · speech-to-text

Sources

2ggml.aiggml.ai · read 2026-10-29
3Free AI Voice Generator & Voice Agents Platform | ElevenLabselevenlabs.io · read 2026-10-29
4LocalAI · Make AI run on every machinelocalai.io · read 2026-10-29
6gguf: declares "gguf" (the record we read and signed)unfragile.ai · read 2026-10-24
7ElevenLabs MCP Server: declares "speech_to_text" (the record we read)github.com · read 2026-10-24
8LocalAI (Activepieces): declares "transcribe_audio" (the record we read and signed)unfragile.ai · read 2026-10-24

For agents: the 10 actions behind this answer

spuuntries/flatdolphinmaid-8x7b-gguf · spuuntries/flatdolphinmaid-8x7b-gguf verified and signed
Undi95's FlatDolphinMaid 8x7B Mixtral Merge, GGUF Q5_K_M quantized by TheBloke.
kcaverly/neuralbeagle14-7b-gguf · kcaverly/neuralbeagle14-7b-gguf verified and signed
NeuralBeagle14-7B is (probably) the best 7B model you can find!
gguf · gguf verified and signed
openai/gpt-5-nano · openai/gpt-5-nano verified and signed
Fastest, most cost-effective GPT-5 model from OpenAI
google/nano-banana-2 · google/nano-banana-2 verified and signed
Google's fast image generation model with conversational editing, multi-image fusion, and character consistency
speech_to_text · ElevenLabs MCP Server found in code, not yet checked
openai/gpt-4.1-nano · openai/gpt-4.1-nano verified and signed
Fastest, most cost-effective GPT-4.1 model from OpenAI
jaaari/kokoro-82m · jaaari/kokoro-82m verified and signed
Kokoro v1.0 - text-to-speech (82M params, based on StyleTTS2)
openai/whisper · openai/whisper verified and signed
Convert speech in audio to text
nvidia/canary-qwen-2.5b · nvidia/canary-qwen-2.5b verified and signed
🎤The best open-source speech-to-text model as of Jul 2025, transcribing audio with record 5.63% WER and enabling AI tasks like summarization directly from speech✨

Labels are the instruments' verdict and are never changed by ranking. Agents and robots get this same answer as JSON at /api/v1/find?q=…&answer=1 and through the MCP server.