FireRedTeam's FireRedASR2 recognition models, FireRedASR2-LLM and FireRedASR2-AED, support two languages, Chinese (Mandarin, with 20+ dialects and accents) and English, plus code-switching.

Quote found word for word on the live source, last checked October 20, 2026

In the source's own words

“Model Supported Languages & Dialects Download FireRedASR2-LLM Chinese (Mandarin and 20+ dialects/accents * ), English, Code-Switching”

FireRedASR2S: README, FireRedTeam, GitHub, FireRedTeam/FireRedASR2S, 2026, Available Models and Languages table.

What it measures: The table gives the same languages for FireRedASR2-LLM and FireRedASR2-AED. The same table lists 100+ languages for FireRedLID (spoken language identification) and FireRedVAD (voice activity detection). Those models identify or detect speech and do not transcribe it, so they are not counted here. The 20+ dialects and accents are Chinese varieties, and code-switching is a mode of speech, so neither is counted as a language.
What it does not show: The 2 is counted from the table, because the README does not state a number. The table is the only support list the README gives, and it names no other languages for the recognition models.

A row of How Many Languages Speech Recognition Supports, as FireRedASR2 (Chinese, English): 2 languages.

System: FireRedASR2S
Maker: FireRedTeam
Kind: Open model
What is counted: Languages
On device: Not stated
As of: 2026-10
Note: Counted from the table: Chinese and English are the two languages. The 20+ Chinese dialects and accents and code-switching are not counted as languages, so the 2 is a count, not a number the page states.

Fact · Languages and accents

2languages

Quote found word for word on the live source, last checked October 20, 2026

“Model Supported Languages & Dialects Download FireRedASR2-LLM Chinese (Mandarin and 20+ dialects/accents * ), English, Code-Switching”

FireRedASR2S: README, FireRedTeam, GitHub, FireRedTeam/FireRedASR2S, 2026, Available Models and Languages table.

What it measures: The table gives the same languages for FireRedASR2-LLM and FireRedASR2-AED. The same table lists 100+ languages for FireRedLID (spoken language identification) and FireRedVAD (voice activity detection). Those models identify or detect speech and do not transcribe it, so they are not counted here. The 20+ dialects and accents are Chinese varieties, and code-switching is a mode of speech, so neither is counted as a language.
What it does not show: The 2 is counted from the table, because the README does not state a number. The table is the only support list the README gives, and it names no other languages for the recognition models.

A row of How Many Languages Speech Recognition Supports, as FireRedASR2 (Chinese, English): 2 languages.

System: FireRedASR2S
Maker: FireRedTeam
Kind: Open model
What is counted: Languages
On device: Not stated
As of: 2026-10
Note: Counted from the table: Chinese and English are the two languages. The 20+ Chinese dialects and accents and code-switching are not counted as languages, so the 2 is a count, not a number the page states.