CallMissed logoCallMissed
Back to all models
LLM Chatindian-languagesfree-tiervoicereasoningtools
Sarvam AI — sarvam-105b-conversations logo

Sarvam 105B Conversations

by Sarvam AI · Released 2026

A 105B Mixture-of-Experts model from Sarvam AI tuned for conversation and voice. 128K context, tool calling, reasoning, and streaming — built for multi-turn dialogue rather than one-shot completion. Available on the CallMissed free tier.

Sarvam AI — sarvam-105b-conversations logo
LLM Chat

Sarvam 105B Conversations

Powered by Sarvam AI · Mixture-of-Experts (MoE)

Context Window

128K

Parameters

105B (MoE)

Max Output

Undisclosed

Category

LLM Chat

Overview

Sarvam 105B Conversations is a conversation-tuned variant of Sarvam AI's 105-billion-parameter Mixture-of-Experts model. Where the base 105B is a general-purpose model, this variant is post-trained for dialogue: holding context across many turns, keeping a consistent persona, and producing the short, speakable responses that voice agents and chat assistants need.

It carries the full 131,072-token context window, so an entire call transcript, knowledge-base excerpt, or long support thread fits in one request. Tool calling lets the model reach into your systems mid-conversation — look up an order, check a booking, escalate a ticket — and reasoning support helps it plan multi-step answers before replying. Streaming is supported end to end, which matters for voice: partial tokens can be handed to a TTS model as they arrive instead of waiting for the full completion.

The model is text-only — it does not accept image input. Pricing is $0.35 per million tokens for both input and output, and it is included in the CallMissed free tier, so you can prototype a voice or chat agent on it before committing to paid capacity.

Pricing

MetricPrice
Input /1M tokens₹35.0000
Output /1M tokens₹35.0000

1 credit = ₹1 = $0.01 USD. Prices shown from provider; CallMissed passes through with ~35% markup.

Key Highlights

  • Tuned for conversation and voice, not one-shot completion
  • 128K context — full call transcripts in one request
  • Tool calling, reasoning, and streaming
  • Free on CallMissed — available on the free tier

Technical Details

  • Architecture: Mixture-of-Experts (MoE) with 105B total parameters
  • Context window: 131,072 tokens
  • Post-trained for multi-turn conversation and voice-agent use
  • Tool calling (function calling) supported
  • Reasoning supported
  • Streaming responses supported — pairs with a TTS leg for voice
  • Text-only — does not accept image input
  • Free tier eligible on CallMissed

Strengths

  • Conversation-tuned — holds context and persona across long dialogues
  • 128K context fits full transcripts and knowledge-base excerpts
  • Tool calling plus reasoning for agents that act, not just answer
  • Streaming output keeps voice latency low
  • Same $0.35/$0.35 pricing as the base 105B, and free-tier eligible

Limitations

  • Text-only — no vision or image input
  • Tuned for dialogue, so less suited to long-form single-shot generation
  • MoE architecture needs more memory at deployment than a dense model with the same active parameter count

Use Cases

Voice agentsMulti-turn chat assistantsCustomer support automationTool-using conversational agents

API Example

curl https://api.callmissed.com/v1/chat/completions \
  -H "Authorization: Bearer cm_YOUR_KEY" \
  -d '{"model": "sarvam-105b-conversations", "messages": [{"role": "user", "content": "Mera order kab tak aayega?"}], "stream": true}'

Endpoint: POST /v1/chat/completions · Model ID: sarvam-105b-conversations

Try Sarvam 105B Conversations now

Get 1000 free API credits on signup. No credit card required.

Start free Book a DemoRead docs