> This page is for Model APIs. > Sarvam Voice Agents (Samvaad) is a voice AI agent platform for Indian languages: build, test and deploy AI calling agents on your own phone numbers, from ₹3 a minute. Start here: https://docs.sarvam.ai/voice-agents/why-voice-agents. > For clean Markdown of any page, append `.md` to the page URL. > For a complete documentation index, see https://docs.sarvam.ai/llms.txt. For a section index, append `/llms.txt` to any section URL (for example https://docs.sarvam.ai/voice-agents/llms.txt). > For AI client integration (Claude Code, Cursor, etc.), connect to the MCP server at https://docs.sarvam.ai/_mcp/server. # DeepSeek V4 Pro > DeepSeek V4 Pro is the larger open-weight DeepSeek V4 reasoning model, with text input and output, a 1,000,000-token context window, and tool calling. DeepSeek V4 Pro is the larger model in the DeepSeek V4 family, served on Sarvam infrastructure as an [open-weight model](/api/token-factory/models). It supports text input and output with a **1,000,000-token context window**. > **Note** > > DeepSeek V4 Pro is in **beta**. Its behaviour, limits, and availability may change before general availability. Beta models must be enabled for your API key; without access, requests return `400 invalid_request_error` ("This model is currently in beta"). [Contact us](/api/getting-started/help) to request access. See [Beta models](/api-reference/beta-apis). ## At a glance | | | | ------------------ | ---------------------------------------------------------------------------------------------------------------------------------- | | **Model ID** | `deepseekv4-pro` | | **Best for** | Harder reasoning and agentic tasks over long text | | **Context window** | 1,000,000 tokens | | **Input → output** | Text → text | | **Reasoning** | **`low`**, **`high`** (default), or **`max`**; can be turned off. See [Reasoning](#reasoning) | | **Tool calling** | Supported, including `tool_choice: "required"` and parallel tool calls | | **APIs** | [`POST /v2/chat/completions`](/api-reference/chat/chat-completions-v2) and [`POST /v2/responses`](/api-reference/responses/create) | | **Pricing** | ₹115 input · ₹4 cached input · ₹350 output per 1M tokens; reasoning is billed as output ([Pricing](/api/getting-started/pricing)) | ## Key capabilities #### Long context Process up to 1,000,000 tokens across the prompt, reasoning, and generated output. #### Reasoning Choose `low`, `high`, or `max`, or turn thinking off for direct answers. #### Tool calling Use OpenAI-compatible `tools` and `tool_choice`, with thinking on or off. #### Structured output Constrain answers with `response_format` JSON schemas. #### Log probabilities Request `logprobs` and `top_logprobs`. #### Multiple choices Generate `n` > 1 completions in one request. > **Tip** > > Moving from [OpenRouter](/api/token-factory/migrations/from-openrouter/overview), [Fireworks AI](/api/token-factory/migrations/from-fireworks/overview), or [Together AI](/api/token-factory/migrations/from-together/overview)? The migration guides map model IDs and request fields to Sarvam. ## Quickstart #### Python ```python from sarvamai import SarvamAI client = SarvamAI(api_subscription_key="YOUR_SARVAM_API_KEY") response = client.chat.completions_v2( model="deepseekv4-pro", messages=[ { "role": "user", "content": "Explain the difference between a B-tree and a B+ tree index.", } ], max_tokens=2000, ) print(response.choices[0].message.content) ``` #### JavaScript ```javascript import { SarvamAIClient } from "sarvamai"; const client = new SarvamAIClient({ apiSubscriptionKey: "YOUR_SARVAM_API_KEY", }); async function main() { const response = await client.chat.completionsV2({ model: "deepseekv4-pro", messages: [ { role: "user", content: "Explain the difference between a B-tree and a B+ tree index.", }, ], max_tokens: 2000, }); console.log(response.choices[0].message.content); } main(); ``` #### cURL ```bash curl -X POST https://api.sarvam.ai/v2/chat/completions \ -H "api-subscription-key: $SARVAM_API_KEY" \ -H "Content-Type: application/json" \ -d '{ "model": "deepseekv4-pro", "messages": [ { "role": "user", "content": "Explain the difference between a B-tree and a B+ tree index." } ], "max_tokens": 2000 }' ``` ## Supported parameters `model: "deepseekv4-pro"` on [Chat Completion V2](/api-reference/chat/chat-completions-v2) and [Responses](/api-reference/responses/create). Full request/response schemas live on those API reference pages; this table covers what's actually accepted for this model. | Parameter | Chat Completion V2 | Responses | | ----------------------------------------------------- | :-----------------------------------------------------------------------: | :---------------------------------------------------------------------------------------------------: | | Streaming (`stream`) | | | | Tool calling (`tools`, `tool_choice`) | | | | `parallel_tool_calls: false` enforced | | | | Structured output (`response_format` / `text.format`) | | | | Log probabilities (`logprobs`, `top_logprobs`) | | send `top_logprobs` with `include: ["message.output_text.logprobs"]`; `logprobs: true` returns `400` | | Multiple choices (`n` > 1) | | N/A — Responses has no `n` | | Reasoning control | `reasoning_effort`: `low`, `high` (default), `max`, or `none` to turn off | `reasoning.effort`: same values | | Turn reasoning off | `reasoning_effort: "none"` | `reasoning.effort: "none"` | > **Warning** > > On `POST /v2/responses`, `logprobs` is only populated when the request includes > `"include": ["message.output_text.logprobs"]`. Without it, `output[].content[].logprobs` > is always an empty array even if `top_logprobs` is set. Verified against the live API on > 2026-10-22. ## Reasoning DeepSeek V4 Pro thinks by default at **`high`** effort. | Value | Behaviour | | ---------------------------------- | ----------------------------------------------------------- | | `"low"`, `"high"`, `"max"` | Sets the reasoning effort. Omitting the field gives `high`. | | `"none"` | Turns thinking off. | | `"minimal"`, `"medium"`, `"xhigh"` | Also accepted. | | Integer budgets | Not supported; return `400`. | Set the value with `reasoning_effort` on [Chat Completion V2](/api-reference/chat/chat-completions-v2) or `reasoning.effort` on [Responses](/api-reference/responses/create). You can also turn thinking off with `thinking: {"type": "disabled"}`, sent through `extra_body` in the SDKs: ```python response = client.chat.completions_v2( model="deepseekv4-pro", messages=[{"role": "user", "content": "What is 17 × 23?"}], extra_body={"thinking": {"type": "disabled"}}, max_tokens=2000, ) ``` Read the reasoning trace from `choices[].message.reasoning_content` on chat. On Responses, `output` starts with a `reasoning` item; read its `content[0].text` (`summary` is empty). Reasoning tokens are billed as output tokens and count against `max_tokens` / `max_output_tokens`. ## Tool calling Pass OpenAI-compatible `tools`. `tool_choice` accepts `"auto"`, `"none"`, `"required"`, or a named function (`{"type": "function", "function": {"name": "..."}}`). Tool calling works with thinking on or off. The model can return several tool calls in one turn; set `parallel_tool_calls: false` to get at most one. ## Known limitations * Text input only. Image, audio, and video parts return `400`. * Reasoning and the final answer share the **`max_tokens`** / **`max_output_tokens`** budget. * `stop` strings apply to the answer only; a match inside the reasoning doesn't end the response. * Structured output (`response_format` on chat, `text.format` on Responses) turns reasoning off, even when you set an effort level. * Set **`max_tokens`** explicitly. When it's omitted, generation isn't capped at a small default and long answers can run to tens of thousands of tokens. ## Limits and errors | Area | Limit or condition | API response | Recommended action | | -------------- | ----------------------------------------------- | ---------------------------------------------------------- | ------------------------------------------------------------------------- | | Output | `max_tokens` exceeds 1,000,000 | `400 invalid_request_error` | Set `max_tokens` within the context window. | | Request size | Request body exceeds 10 MiB (10,485,760 bytes) | `413 invalid_request_error`, with the limit in the message | Shorten the prompt. | | Modality | Request includes image, audio, or video input | `400 invalid_request_error` | Send text input only. | | Reasoning | Reasoning consumes the full `max_tokens` budget | `200` with `finish_reason: "length"` | Raise `max_tokens` or lower the reasoning effort. | | Authentication | API key is missing or invalid | `403 invalid_api_key_error` | Check the `api-subscription-key` header. | | Model | Model ID is unknown or unavailable | `404 not_found_error` | Check the spelling; model IDs are case-sensitive. | | Quota | No credits remain | `402 insufficient_quota_error` | Add credits in the [dashboard](https://indus.sarvam.ai/billing?ref=docs). | | Rate limit | Request or concurrency limit is exceeded | `429 rate_limit_exceeded_error` | Back off and retry. | | Availability | Model remains overloaded after retries | `503 model_overloaded` | Retry later. |