Indic Models

LLM

sarvam-105b

• Architecture: 105B MoE, MLA architecture
• Context: 128K tokens
• Training: Pre-trained on 12T tokens
• Best for: Complex reasoning, agentic tasks, long documents
• Thinking mode: reasoning_effort: "low" | "medium" | "high"

sarvam-105b-conversations

• Architecture: 105B MoE, tuned for conversation and voice
• Context: 32K tokens
• Tool calling: yes
• Streaming: yes
• Best for: Multi-turn dialogue, voice agents, assistants that talk
• Thinking mode: reasoning_effort: "low" | "medium" | "high"

Same family and same price as sarvam-105b, with a shorter 32K window — tuned for spoken dialogue rather than long-form work. It does not accept image input.

Thinking Mode

The sarvam-105b models support hybrid thinking mode:

response = client.chat.completions.create(
    model="sarvam-105b",
    messages=[{"role": "user", "content": "Solve this complex problem step by step"}],
    extra_body={"reasoning_effort": "high"}
)

"medium" is not one of the models' levels, so it is dropped and the model runs at its default. Thinking tokens count toward max_tokens; when you omit it we send 4,096 so the answer is not cut off by the upstream default of 2,048. See the reasoning_effort matrix.

Speech to Text

saaras:v3

• Languages: 23 (22 Indic + English)
• Output modes: transcribe, translate, verbatim, translit, codemix
• Auto language detection: yes
• Telephony support: 8kHz audio
• Endpoint: POST /v1/audio/transcriptions

Supported languages include: Hindi, Bengali, Gujarati, Kannada, Malayalam, Marathi, Odia, Punjabi, Tamil, Telugu, Urdu, Assamese, Bodo, Dogri, Kashmiri, Konkani, Maithili, Manipuri, Nepali, Sanskrit, Santali, Sindhi, and English.

saaras:v4

• Languages: 24
• Output modes: transcribe, translate, verbatim, translit, codemix
• Auto language detection: yes
• Endpoint: POST /v1/audio/transcriptions

Five output modes on one model — standard transcription, English translation, verbatim (fillers kept), Latin-script transliteration, and code-mixed output. Select one with the mode form field; transcribe is the default.

Text to Speech

bulbul:v3

• Voices: 37 speakers
• Languages: 11
• Audio codecs: WAV, MP3, OPUS, FLAC, AAC, Mulaw, Alaw, PCM
• Pace: 0.5–2.0 (maps to speed parameter)
• Sample rates: 8000, 16000, 22050, 24000, 48000 Hz
• Endpoint: POST /v1/audio/speech

Default voice: shubh. See the Voices page for the full list.