DataAIHub
DataAIHubNews · Research · Tools · Learning
SarvamOpen SourceReasoningSmall Models

Sarvam-M

Sarvam’s earlier multilingual model: a 24B hybrid-reasoning post-train of Mistral Small, released under Apache-2.0 in May 2025. API id sarvam-m.

Thinking · Coding

Last reviewed: 25 September 2026

Official pricing →

Overview

Sarvam’s earlier multilingual model: a 24B hybrid-reasoning post-train of Mistral Small, released under Apache-2.0 in May 2025. API id sarvam-m.

Capabilities

  • Vision: No
  • Audio: No
  • Tool calling: No
  • Thinking: Yes
  • MCP: No
  • Coding: Yes
  • Structured output: No

Technical specifications

Provider
Sarvam
License
Apache-2.0
Parameters
24B
Architecture
Post-train of Mistral-Small-3.1-24B-Base
Release
2025-05-23
Modalities
Text
Vision
No
Audio
No
Tool calling
No
Thinking
Yes
MCP
No
Open weights
Yes
API
Yes
Pricing (input)
Sarvam API (model id sarvam-m)
Pricing (output)
Sarvam API (model id sarvam-m)

Supported modalities

Text

Context window

Not specified

Pricing

Input: Sarvam API (model id sarvam-m)
Output: Sarvam API (model id sarvam-m)

reasoning_effort of low, medium, or high enables thinking mode on the OpenAI-compatible API. Self-host with vLLM 0.8.5 or newer.

Availability

API: Yes
Chat UI: No
Open weights: Yes

Prefer Sarvam 30B or 105B for new deployments unless you already serve sarvam-m.

Use cases

  • Multilingual Indian-language chat
  • Self-hosted Mistral-Small-class weights
  • Existing sarvam-m API integrations

Strengths

  • Apache-2.0 and a familiar Mistral Small base
  • OpenAI-compatible API with a thinking mode
  • Covers major Indian languages plus English

Limitations

  • Superseded for most new work by the from-scratch 30B and 105B models
  • Text-only

Related guides

Related benchmarks

Related research

Related GitHub

Related tools

Related rankings

Companies

Explore more models

All models →