August 2026
Speech to Text (STT)
Realtime API: Introduced a new Realtime API - saaras:v3-realtime over WebSocket - purpose-built for voice agents and live transcription. It adds true partial (interim) transcripts as the user speaks, live mid-stream reconfiguration, and millisecond-based VAD tuning. This supersedes the legacy streaming WebSocket for new voice-agent and live-transcription work; the older saaras:v3 streaming WebSocket remains generally available and is now documented under Legacy.
saaras:v4-realtime Generally Available: saaras:v4-realtime is now available on the Realtime API alongside saaras:v3-realtime, with the same connection parameters and message protocol. saaras:v3-realtime remains the default.
Open-Source Models
GLM-5.2 and Gemma 4 31B (Beta): Introduced two open-source chat models served on Sarvam infrastructure via the new /v2/chat/completions endpoint. GLM-5.2 (glm5.2) offers a 512K-token context window with tool calling and visible reasoning, for tasks that exceed Sarvam’s own 128K context. Gemma 4 31B (gemma4) adds image input alongside tool calling. Neither model is tuned for Indian languages - for Indian-language workloads, continue using Sarvam-105B or Sarvam Vision. Beta access is granted per API key; contact us to request whitelisting.
Self-Hosted Deployments
Self-Hosted Deployments: Sarvam models can now run inside your own AWS account via Self-Hosted Deployments. Subscribe to a model package on the AWS Marketplace and deploy Saaras v3, Bulbul v3, or Sarvam Vision as an Amazon SageMaker endpoint in your own VPC - same models as the Managed API, but your audio and documents never leave your infrastructure.
Pricing
LLM Pricing Updated: sarvam-105b pricing increased to ₹29.28 / ₹10.98 / ₹73.2 per 1M tokens (input / cached input / output), up from ₹4 / ₹2.5 / ₹16, effective August 5, 2026. sarvam-105b-conversations is priced the same as sarvam-105b. Pricing was also published for the beta open-source models - GLM-5.2 at ₹128.1 / ₹23.79 / ₹402.6 and Gemma 4 31B at ₹36.6 / ₹13.73 / ₹91.5 - with GLM-5.2’s reasoning tokens billed as output. See the pricing page for details.
Voice Agents Updates
Renamed from Conversations: The product formerly called Conversations (internally, “Samvaad”) is now Voice Agents, an end-to-end stack for building, deploying, and monitoring voice agents on telephony. References to platform.sarvam.ai were updated accordingly.
Agent Analytics Generally Available: The new Agent Analytics view is now generally available. Open Monitor → Agent Analytics for a period-level health check of your voice agents across eight tabs - Overview, Connectivity, Engagement, Tools, Goals, Group by, Call Logs, and Log Analyser:
- Overview - KPI tiles (calls attempted, pick-up rate, latency, average duration, total billable minutes, short calls, goal achievement), a trend chart with day/week/month/quarter granularity, a call-outcomes donut, top failure reasons, and an agent/campaign rollup table.
- Tools - tool error tracking: six tiles (calls, tools invoked, errors, error rate, p50, p95), an error-type breakdown ranked by frequency, and a per-tool table that flags anything above a ~5% error rate.
- Goals - one goal per agent (voice only): calls scored passed/failed/unevaluated, average turns to goal, and conversion insights broken down by language, time of day, TTS voice, and intro message.
- Call Logs - every call with
attempted_attimestamps, call-forwarding analytics, and turn-level drill-down via Log Analyser.
See Agent Analytics.
Streaming Cohorts (JSON Cohort Uploads): Cohort uploads now support JSON via the Stream cohort API, for teams that want to add contacts programmatically instead of uploading a CSV in the dashboard. Send up to 1,000 users per request, and add contacts to a campaign while it’s running - they queue to dial within the current schedule. See Cohorts.
Code Tools: Added Code Tools, letting you upload a custom Python file to define tools that go beyond the built-in API and data-validation tools. Available on request - contact us to get access.
Call-Context Variables in API Tools: API tools can now receive live call context as variables. In the tool editor, select User Identifier to pass the caller’s phone number, and Call Transcript, Interaction ID, and Call Length to pass the full conversation transcript, the unique interaction ID, and the call’s duration. These are available alongside agent variables - visible in the platform UI and fetchable via the APIs.
Max Call Length Raised to 60 Minutes: The Max call length limit in conversation settings was raised from 25 to 60 minutes, unlocking longer use cases like employee training, guided walkthroughs, and extended support conversations. When the limit is reached, the agent still wraps up and hangs up.
Outbound Webhooks: Campaigns can now POST an event to your webhook endpoint as each call finishes, whether or not the call connected. Set the Webhook URL in the campaign wizard’s Advanced step. The payload carries call outcome, connectivity status, retry metadata, timestamps, and output agent variables. Every delivery is recorded with a status and can be re-delivered via Retry webhook deliveries. See Webhooks under Campaigns.
Content Studio Updates
Renamed from Content Agents: The product formerly called Content Agents is now Content Studio. Page URLs are unchanged.
Dubbing Restructured: Reorganized Dubbing docs into a hub with dedicated Voices and Job Lifecycle pages, and documented editor_flow pricing: ₹40 / ₹38 / ₹36 per minute (Starter / Pro / Enterprise) with editor_flow: false, and ₹80 / ₹75 / ₹72 per minute - 2× the base rate - with editor_flow: true.
Cowork Updates
Renamed from Work Agents: The product formerly called Work Agents is now Cowork. Its docs moved from /work-agents/* to /cowork/*; the previous URLs redirect to the new ones.
Integrations
Vobiz: Added a Vobiz telephony integration guide for Voice Agents.