--- title: Index-Translate-9B-Joint-f35b68981f27-GGUF canonical_url: "https://www.modelscope.ai/models/IndexTeam/Index-Translate-9B-Joint-f35b68981f27-GGUF" md_url: "https://www.modelscope.ai/models/IndexTeam/Index-Translate-9B-Joint-f35b68981f27-GGUF.md" repository: IndexTeam/Index-Translate-9B-Joint-f35b68981f27-GGUF last_updated: 2026-11-07 license: apache-2.0 library_name: - gguf - pytorch frameworks: - pytorch language: - zh - en downloads: 21 stars: 0 tags: - index-translate - multimodal - translation - image-to-text - automatic-speech-recognition - gguf --- # Index-Translate-9B-Joint-f35b68981f27-GGUF > Index-Translate-9B-Joint-f35b68981f27-GGUF - An open-source model by IndexTeam on ModelScope. Index-Translate-omni-9B-GGUF IndexTeam/Index-Translate-9B-Joint-f35b68981f27-GGUF is a machine learning model on ModelScope. licensed under apache-2.0. - **Repository**: IndexTeam/Index-Translate-9B-Joint-f35b68981f27-GGUF - **License**: apache-2.0 - **Tags**: index-translate, multimodal, translation, image-to-text, automatic-speech-recognition, gguf - **Downloads**: 21 - **Stars**: 0 - **Last updated**: 2026-11-07 Source: https://www.modelscope.ai/models/IndexTeam/Index-Translate-9B-Joint-f35b68981f27-GGUF --- # Index-Translate-omni-9B-GGUF Built on the Index-Translate family, with multilingual text translation, terminology and formatting constraints, image translation, speech recognition, and bidirectional Chinese–English speech translation. ## Original Model Reference Source: [Index-Translate-Omni model card](https://huggingface.co/IndexTeam/Index-Translate-omni-9B). The tables below reproduce the upstream results for the original models. They are not new benchmark results for this quantized export; see the format-specific verification below for the checks actually performed here. ## Text Translation | Model | FLORES COMET22 ↑ | WMT24++ COMET22 ↑ | WMT26 Judge ↑ | MEME ↑ | |---|---:|---:|---:|---:| | Index-Translate 2B | 0.8655 | 0.8489 | 60.26 | 0.6443 | | Index-Translate-Omni 2B | 0.8655 | 0.8489 | 60.26 | 0.6443 | | Index-Translate 9B | 0.8789 | 0.8601 | 75.35 | 0.7387 | | Index-Translate-Omni 9B | 0.8789 | 0.8601 | 75.35 | 0.7387 | | Index-Translate 35BA3B (preview) | 0.8794 | 0.8586 | 76.76 | 0.7405 | | Index-Translate-Omni 35BA3B | 0.8795 | 0.8593 | 78.01 | 0.7405 | ### Instruction Following | Model | instTrans Quality / IFscore ↑ | instTrans_minor Quality / IFscore ↑ / Off-target ↓ | IFMTBench XCOMET-XXL / IFscore ↑ | |---|---:|---:|---:| | Index-Translate 2B | 0.5391 / 0.7569 | 0.3050 / 0.6586 / 3.97% | 0.7712 / 0.7584 | | Index-Translate-Omni 2B | 0.5391 / 0.7569 | 0.3050 / 0.6586 / 3.97% | 0.7712 / 0.7584 | | Index-Translate 9B | 0.6771 / 0.8209 | 0.5222 / 0.7725 / 3.47% | 0.7957 / 0.8760 | | Index-Translate-Omni 9B | 0.6771 / 0.8209 | 0.5222 / 0.7725 / 3.47% | 0.7957 / 0.8760 | | Index-Translate 35BA3B (preview) | 0.6901 / 0.8336 | 0.5151 / 0.7715 / 4.05% | 0.7926 / 0.8991 | | Index-Translate-Omni 35BA3B | 0.6867 / 0.8371 | 0.5163 / 0.7843 / 3.62% | 0.7934 / 0.9070 | Original Index-Translate figures come from the [published benchmarks](https://huggingface.co/IndexTeam/Index-Translate-9B#evaluation). The original 35BA3B reference is the published preview checkpoint. ## Speech CER is a percentage (lower is better). BLEU and chrF2 are higher-is-better. Translation cells show BLEU / chrF2 / target CER. | Model | Chinese ASR CER ↓ | English ASR CER ↓ | English → Chinese | Chinese → English | |---|---:|---:|---:|---:| | Index-Translate-Omni 2B | 13.8113 | 34.2560 | 36.982 / 30.357 / 43.3452 | 30.747 / 51.152 / 48.0670 | | Index-Translate-Omni 9B | 17.8917 | 14.4097 | 43.507 / 36.312 / 41.3767 | 34.656 / 55.376 / 46.3720 | | Index-Translate-Omni 35BA3B | 18.6803 | 15.0850 | 42.081 / 35.836 / 49.2490 | 36.404 / 57.323 / 45.1546 | 35BA3B uses a separate frontend for Chinese ASR; English ASR and speech translation use the native profile. Evaluation sample counts differ across model sizes. ## Image Translation COMET22 × 100 (higher is better). Each Omni model is compared with its Qwen3.5 baseline on paired evaluated subsets. | Model | MMTIT ↑ | FLORES-rendered ↑ | Vistra ↑ | |---|---:|---:|---:| | Qwen3.5-2B | 52.164 | 50.187 | 55.326 | | Index-Translate-Omni 2B | 65.809 | 53.612 | 65.596 | | Qwen3.5-9B | 73.814 | 67.804 | 68.963 | | Index-Translate-Omni 9B | 77.604 | 68.438 | 71.815 | | Qwen3.5-35B-A3B | 75.602 | 70.005 | 71.067 | | Index-Translate-Omni 35BA3B | 77.011 | 69.891 | 71.482 | 35BA3B reuses the verified image scores for the same weights. Results cover the evaluated subsets; the original failure and UNKNOWN scoring rules are retained. ## GGUF Export Omni translation model exported for **llama.cpp**, with language weights and a separate vision projector. **Audio limitation:** GGUF contains the language model and optional `mmproj` vision support. It does **not** contain the original audio encoder, audio connector or audio runtime. It is not an end-to-end audio translation model. Use the FP8 or FP4 repository for the portable audio modules. ## Formats | Quantization | File | Size | PPL | Change vs f16 | |---|---|---:|---:|---:| | `f16` | `Index-Translate-omni-9B.f16.gguf` | 17.92 GB | 11.4870 | baseline | | `Q8_0` | `Index-Translate-omni-9B.Q8_0.gguf` | 9.53 GB | 11.5140 | +0.23% | | `Q6_K` | `Index-Translate-omni-9B.Q6_K.gguf` | 7.36 GB | 11.4985 | +0.10% | | `Q5_K_M` | `Index-Translate-omni-9B.Q5_K_M.gguf` | 6.47 GB | 11.7721 | +2.48% | | `Q5_K_S` | `Index-Translate-omni-9B.Q5_K_S.gguf` | 6.31 GB | 11.5769 | +0.78% | | `Q4_K_M` | `Index-Translate-omni-9B.Q4_K_M.gguf` | 5.63 GB | 11.0747 | -3.59% | | `Q4_K_S` | `Index-Translate-omni-9B.Q4_K_S.gguf` | 5.35 GB | 10.3659 | -9.76% | | `Q3_K_L` | `Index-Translate-omni-9B.Q3_K_L.gguf` | 4.93 GB | 12.6886 | +10.46% | | `Q3_K_M` | `Index-Translate-omni-9B.Q3_K_M.gguf` | 4.62 GB | 11.8177 | +2.88% | | `Q3_K_S` | `Index-Translate-omni-9B.Q3_K_S.gguf` | 4.26 GB | 14.4046 | +25.40% | | `Q2_K` | `Index-Translate-omni-9B.Q2_K.gguf` | 3.83 GB | 27.3589 | +138.17% | | `IQ4_XS` | `Index-Translate-omni-9B.IQ4_XS.gguf` | 5.23 GB | 11.3353 | -1.32% | `f16` is the conversion baseline, not an FP8 file. The `mmproj-f16` and `mmproj-Q8_0` files are vision projectors, not standalone language models. For 35B, download all F16 shards when using F16; the other quantizations are single-file GGUFs. ## Usage Use a llama.cpp build at or after commit `9e2a04a439edc49f08c5acdafefae19dfa25a658` with Qwen3.5 support: ```bash llama-cli -m Index-Translate-omni-9B.Q4_K_M.gguf -ngl 99 --single-turn \ -p 'Translate into English: 人工智能正在改变内容创作。' ``` For image input, add `--mmproj Index-Translate-omni-9B.mmproj-f16.gguf --image /path/to/image.jpg`. Download the language GGUF and the projector separately. ## Verification Every listed language quantization and both vision projectors were loaded and exercised on an H100 using llama.cpp. Language perplexity uses the same short held-out corpus, 128-token context and one chunk. The table is a small regression check, not a comprehensive translation benchmark; differences on this sample must not be interpreted as a quality ranking. Both projectors were tested with an image-generation call. End-to-end audio was not tested for GGUF because it is outside this format. See `validation.json` and `SHA256SUMS.txt` for exact results and file hashes. ## Mirrors - Hugging Face: https://huggingface.co/IndexTeam/Index-Translate-omni-9B-GGUF - ModelScope: https://modelscope.cn/models/IndexTeam/Index-Translate-omni-9B-GGUF