CoVoST2
Nova 2 Omni Preview 40.7 leads since Dec 2025 Fading Models translate speech from 21 languages into English and from English into 15 languages. Scored by BLEU.
In the authors’ words
Speech translation has recently become an increasingly popular topic of research, partly due to the development of benchmark datasets. Nevertheless, current datasets cover a limited number of languages. With the aim to foster research in massive multilingual speech translation and speech translation for low resource language pairs, we release CoVoST 2, a large-scale multilingual speech translation corpus covering translations from 21 languages into English and from English into 15 languages. This represents the largest open dataset available to date from total volume and language coverage perspective. Data sanity checks provide evidence about the quality of the data, which is released under CC0 license. We also provide extensive speech recognition, bilingual and multilingual machine translation and speech translation baselines with open-source implementation.
- Level
- University Undergraduates and entry-level professionals
- Scored
- Checked automatically A program checks the result against an answer key, unit tests or the expected final state.
- Answers
- Open-ended Free-form text, documents, designs or media judged as a whole.
Nova 2 Omni Preview leads with 40.7
Best score per model as labs report them in their own cards. Labs pick their own settings, so these compare less cleanly than independent runs.
- 1 Nova 2 Omni Preview 40.7
- 2
Gemini 2.0 Pro 40.6 - 3
Phi-4-multimodal 39.3 - 4 Gemini 2.0 Flash Dec 2024 39.2
- 5 Gemini 2.0 Flash Feb 2025 39.0
- 6 Gemini 2.5 Pro Jun 2025 38.5
- 7 Gemini 2.0 Flash-Lite 38.4
- 8
LongCat-Flash-Omni 37.3 - 9 Gemini 2.5 Flash Jun 2025 36.1
- 10
Nemotron 3.5 Bolt E4B 36.1 - 11
Qwen2.5-Omni-7B 34.7
| # | Model | Score | Setting | Date | Source |
|---|---|---|---|---|---|
| 1 | Nova 2 Omni Preview | 40.7 | Dec 2025 | Model card | |
| 2 | Gemini 2.0 Pro | 40.6 | Feb 2025 | Model card | |
| 3 | Phi-4-multimodal | 39.3 | Mar 2025 | Model card | |
| 4 | Gemini 2.0 Flash Dec 2024 | 39.2 | Dec 2024 | Model card | |
| 5 | Gemini 2.0 Flash Feb 2025 | 39.0 | Feb 2025 | Model card | |
| 6 | Gemini 2.5 Pro Jun 2025 | 38.5 | Jul 2025 | Model card | |
| 7 | Gemini 2.0 Flash-Lite | 38.4 | Feb 2025 | Model card | |
| 8 | LongCat-Flash-Omni | 37.3 | Oct 2025 | Model card | |
| 9 | Gemini 2.5 Flash Jun 2025 | 36.1 | Jul 2025 | Model card | |
| 10 | Nemotron 3.5 Bolt E4B | 36.1 | Oct 2026 | Model card | |
| 11 | Qwen2.5-Omni-7B | 34.7 | Mar 2025 | Model card |
3 models have held number one since 2023
Lab claims count from the day the card came out, independent runs from the model’s release.
Scores over time
All the scores we have, one dot each, so a model tested at several settings appears more than once.
Labs headline it less often now, last in Apr 2026
Launch posts from every lab that has listed it, one tick per post. Tall colored ticks include it, and short gray ones leave it out.
| Date | Launch post | Score | As the post writes it |
|---|---|---|---|
| 2 Apr 2026 | Gemma 4 31B IT | CoVoST (31B: no score; audio only on small models) | |
| 2 Dec 2025 | Nova 2 Omni Preview | 40.7 | CoVoST 2 BLEU (Spanish, French, German, Italian to English) |
| 31 Oct 2025 | LongCat-Flash-Omni | 37.3 | Mean of CoVost2 en→zh 47.23 / CoVost2 zh→en 27.32 |
| 27 Mar 2025 | Qwen2.5-Omni-7B | 34.7 | Mean of CoVoST2 en-de 30.2 / de-en 37.7 / en-zh 41.4 / zh-en 29.4 |
| 5 Feb 2025 | Gemini 2.0 Flash-Lite | 38.4 | CoVoST2 (21 lang), BLEU |
| 5 Feb 2025 | Gemini 2.0 Flash | 39 | CoVoST2 (21 lang), BLEU |
| 5 Feb 2025 | Gemini 2.0 Pro | 40.6 | CoVoST2 (21 lang), BLEU |
| 11 Dec 2024 | Gemini 2.0 Flash | 39.2 | CoVoST2 (21 lang) |
9 system cards report it, from 6 labs
Besides launch posts, we read the system cards, model cards and technical reports labs publish with a release. They report far more, including safety and frontier-risk tests.
| Date | Document | Score | As the document writes it | Section |
|---|---|---|---|---|
| 23 Oct 2026 | Nemotron 3.5 Bolt E4B technical report | 36.1 | CoVoST (BLEU), excluding Chinese, reasoning, 37.61 non-reasoning (Table 8) | Capabilities |
| 2 Dec 2025 | Nova 2 Omni Preview technical report | 40.7 | CoVoST 2 BLEU, Spanish, French, German and Italian to English (Table 7) | Capabilities |
| 31 Oct 2025 | LongCat-Flash-Omni technical report | CoVost2 en→zh 47.23, zh→en 27.32, BLEU, no overall score (Table 9) | Capabilities | |
| 7 Jul 2025 | Gemini 2.5 Pro technical report | 38.5 | CoVoST2, 21 languages to English, BLEU (Table 5) | Capabilities |
| 7 Jul 2025 | Gemini 2.5 Flash technical report | 36.1 | CoVoST2, 21 languages to English, BLEU (Table 5) | Capabilities |
| 15 Apr 2025 | Gemini 2.0 Flash model card | 39 | CoVoST2 (21 lang), BLEU (Table, p.3, column Gemini 2.0 Flash (GA)) | Capabilities |
| 10 Apr 2025 | Gemini 2.0 Flash-Lite model card | 38.4 | CoVoST2 (21 lang), BLEU (Table, p.3) | Capabilities |
| 26 Mar 2025 | Qwen2.5-Omni-7B technical report | CoVoST2 BLEU en-de 30.2, de-en 37.7, en-zh 41.4, zh-en 29.4, no overall score (Table 2) | Capabilities | |
| 3 Mar 2025 | Phi-4-multimodal technical report | 39.3 | CoVoST2 X-EN BLEU, 0-shot, 40.76 with CoT. EN-X 37.82, 38.73 with CoT (Tables 3 and 5) | Capabilities |
More in Audio and video
- Video-MMMU In use 300 expert-level lecture videos with questions testing perception, comprehension and adaptation of knowledge. Accuracy. Qwen 3.8 Max 88.7%
- LVBench In use Hours-long video understanding with multiple-choice questions on reasoning, grounding and summarization. Gemini 4 Argon 91.7%
- MMVU In use 3,000 expert-annotated questions on videos across 27 disciplines. Multiple-choice and open-ended accuracy. Qwen 3.8 Max 82.4%
- MLVU In use Multi-task long video understanding (minutes to hours) with multiple-choice and generation tasks. Accuracy. Qwen 3.8 Max 90.8%
- VoiceBench In use Spoken instructions with speaker, environment and content variation. Voice assistant responses scored by GPT judge or rules. Qwen3.5-Omni-Plus 93.1%
- TVBench In use Models answer multiple-choice questions on 10 temporal video tasks, such as action counting and ordering, that static frames cannot solve. Scored by accuracy. Qwen 3.8 Max 81.9%