Codesota · Models2,268 models indexed · 25 match filter
Editorial · Models
Models with recorded evidence.
Start with a research area, drill into a vendor, or page through the full index. Vendor aliases and legacy area IDs are grouped here; original model IDs and model links stay unchanged. Only models with at least one benchmark score appear — a model without a recorded score can’t be ranked.
Vendor:Areas overviewSpeakLeash · 263Alibaba · 104Google · 102OpenAI · 86Meta · 68Microsoft · 49Anthropic · 44DeepSeek · 34Mistral · 30mistralai · 19CYFRAGOVPL · 14NVIDIA · 14Zhipu AI · 13internlm · 10xAI · 10ByteDance · 9Baidu · 8ibm-granite · 8PLLuM · 8allenai · 7Amazon · 7MiniMax · 7Mistral AI · 7Remek · 7Shanghai AI Lab · 7utter-project · 7CohereForAI · 6Salesforce · 601-ai · 5Cohere · 5Moonshot AI · 5NousResearch · 5THUML · 5gguf-iq · 4IBM · 4Meituan · 4openchat · 4Stanford · 4THUDM · 4tiiuae · 4UC San Diego · 4VikParuchuri · 4Allen AI · 3BAAI · 3Du et al. · 3ForgeCode · 3Fudan University · 3gguf · 3gguf11bv30 · 3gguf7bv30 · 3IDEA Research · 3Liao et al. · 3Moonshot.AI · 3Nam Tuan Ly / NII · 3OpenDataLab · 3OPI-PG · 3upstage · 3ViCoS Lab Ljubljana · 3Xiaomi · 3Zhao et al. · 3+ 243 smaller vendors (288 models)
§ 01 · Audio models
25 models in Audio · page 1 of 1.
| # | Model | Vendor | Parameters | Architecture | Benchmarks | Results |
|---|---|---|---|---|---|---|
| 001 | Audio Flamingo 3 | — | — | — | 7 | 7 |
| 002 | CLAP (HTSAT-RoBERTa, fusion, K2C Aug.; T->A) | — | — | — | 1 | 3 |
| 003 | StyleTTS 2 | Columbia University | N/A | TTS | 2 | 2 |
| 004 | VALL-E 2 | Microsoft | Unknown | Neural codec language model (EnCodec tokens) | 2 | 2 |
| 005 | VITS | Kakao | N/A | TTS | 2 | 2 |
| 006 | Voicebox | Meta | 330M | Flow matching (non-autoregressive) | 2 | 2 |
| 007 | XTTS v2 | Coqui AI | Unknown | GPT-based TTS with VQ-VAE tokens | 2 | 2 |
| 008 | AST (Ensemble-M) | — | — | — | 1 | 1 |
| 009 | AST-P | — | — | — | 1 | 1 |
| 010 | AST-S | — | — | — | 1 | 1 |
| 011 | ATST-SED | Audio Research | ~90M | Audio Teacher-Student Transformer for SED | 1 | 1 |
| 012 | AudioCaps baseline (TopDown+Align) | Kim et al. | Unknown | VGGish + Top-Down attention + alignment loss | 1 | 1 |
| 013 | BEATs (iter3+) | — | — | — | 1 | 1 |
| 014 | CLAP+K2C Aug. | — | — | — | 1 | 1 |
| 015 | EnCLAP-base | KAIST / NAVER | ~200M | EnCodec + CLAP encoders + BART decoder | 1 | 1 |
| 016 | FDY-CRNN | KAIST | ~11M | Frequency Dynamic Convolution CRNN | 1 | 1 |
| 017 | MusicGen Large | Meta | 3.3B | Single-stage transformer LM over EnCodec tokens | 1 | 1 |
| 018 | MusicLM | ~1.5B | Hierarchical autoregressive (SoundStream+w2v-BERT+MuLan) | 1 | 1 | |
| 019 | NaturalSpeech 3 | Microsoft | ~500M | Factorized codec + non-AR diffusion | 1 | 1 |
| 020 | NaturalSpeech 3 | Microsoft | Unknown | Discrete diffusion + codec LM (factored) | 1 | 1 |
| 021 | Noise2Music | Unknown | Cascaded diffusion (waveform + spectrogram) | 1 | 1 | |
| 022 | Pengi | Microsoft | ~300M | CLAP audio encoder + GPT-2 decoder | 1 | 1 |
| 023 | Prototypical CRNN (DCASE2021 baseline) | DCASE | ~1.5M | CRNN with mean-teacher | 1 | 1 |
| 024 | VALL-E | Microsoft | ~400M | Neural codec LM (EnCodec tokens) | 1 | 1 |
| 025 | YourTTS | Edresson Casanova et al. | Unknown | VITS-based + speaker conditioning | 1 | 1 |