Tasks and inputs
Sarvam-M is listed for Text generation. It accepts text and produces text. Use the publisher’s input format and preprocessing for your chosen deployment.
Deprecated
Sarvam has retired this model from its API.
Sarvam-M from Sarvam: Hybrid-reasoning multilingual model for Indian languages built on Mistral Small.
The specification score summarizes recorded features and limits. It is not a benchmark result.
Recorded modality support shows what this model can accept and produce. An undocumented modality is shown as unknown rather than assumed to be unsupported.
Accepts
Produces
Hybrid-reasoning multilingual model for Indian languages built on Mistral Small. The registry places it in the sarvam family with a release date of 2025-05-20. This profile describes its documented interface and how to evaluate it for the intended task.
The recorded inputs are text, with text output. The registry lists a 32,768-token context window and up to 32,768 output tokens. Additional capability flags are not established by this record. These are specifications, not a measured quality or speed ranking. A host can impose its own request limits and support a different subset of features.
The registry marks Sarvam-M as an open-weight model. Check the linked checkpoint’s licence and runtime requirements before downloading or deploying it. Open weights describe availability, while licence terms determine permitted use. Local inference still requires hardware; a hosted route can charge separately.
No independently checked token rate is added from a hosting alias. Use the route’s current price and billing unit when comparing cost, including output, reasoning, cache behavior and retries where applicable.
Start with representative prompts and an agreed acceptance rubric. Include missing information, contradictory evidence and outputs that must follow a schema. Compare accepted results, complete request cost and response time under the same deployment conditions.
The publisher has not yet confirmed every specification above in a model-specific document, so check access, limits and pricing on your chosen route before an integration decision. SyncDev has not independently benchmarked Sarvam-M.
Browse Sarvam models, explore related tasks, or compare two models using the same workload.
Compare the model on the work you need to complete.
Sarvam-M is listed for Text generation. It accepts text and produces text. Use the publisher’s input format and preprocessing for your chosen deployment.
Documented limits: 32.77K context and 32.77K maximum output. Measure a completed task using the prompts, documents and tools your application will use. Include reasoning and retries in the result.
The weights are downloadable under Apache 2.0. Check runtime support and size the complete model and working context for your expected concurrency.
These records share a directory category. Check the source card, output type and deployment requirements before treating a model as a replacement.
Microsoft
Aion 1.0 Instruct 2.3B is Microsoft's MIT-licensed on-device small language model and Phi Silica successor for Windows, with a 32K context.
Gemini 4 Neon is Google's mid-tier Gemini 4 model for complex coding and analysis, sitting between Hydrogen and Carbon at $3/$15 per 1M tokens.
OpenAI
GPT-6.2 Astra is OpenAI's November 2026 flagship for hard professional, scientific and agentic work, at $8/$40 per million tokens with a 1.05M context.
OpenAI
GPT-6.2 Sol is OpenAI's everyday model for coding, agents and writing: $2/$10 per million tokens, 1.05M context, released 17 November 2026.
Alibaba
Qwen4 Max (qwen4-max) is Alibaba's first Qwen4-generation flagship: 1M-token context, text, image and video input, thinking on by default, $1.50/$4.50 per 1M.
Sarvam
Sarvam-105B-1711 is a refreshed open post-train of Sarvam's 105B MoE, tuned for tool calling, instruction following, agents and Indian languages.