Language modelEnterprise access

Mercury Voice

Inception

Released: September 29, 2026
Data date: October 3, 2026

Mercury Voice targets the time to first answer token in voice agents. In its own tests on customer-service prompts, Inception reports a median of 320 milliseconds after reasoning. Despite its name, this diffusion LLM generates text; other components handle speech recognition and audio output.

Inception offers the checkpoint through its own API for enterprise customers. It handles 128K context tokens and up to 50K output tokens.

Specifications and access

Specification	Value and source
Model class	Diffusion LLM for voice agentsSource
Context window	128K tokens according to InceptionSource
Maximum output	50K tokens according to InceptionSource
Output	Text, not text-to-speech audioSource
Access	Inception API for enterprise customersSource

Measurements without matching peer values

These measurements have no matching peer values under the same test conditions. Their original values and sources remain available here.

average quality (Show measurement, test conditions, and source)

Source value: 70.5
Score: 70.5
Metric: average quality reported score
Unit: No unit provided
Category: source-specific
Direction: Higher is better
Source type: vendor-reported
Evaluator: inceptionlabs.ai
Status: active
Retrieved at: 2026-10-04
Methodology: Vendor-reported result. Values read from the official image-only table. Quality composite and latency use Inception's voice-agent prompts/configuration.
Context: Source-specific observation; it is not a shared comparison cohort. The source table does not state a unit or scale for this numeric score.

Source inceptionlabs.ai

tau3 Telecom (Show measurement, test conditions, and source)

Source value: 77.2
Score: 77.2
Metric: tau3 Telecom reported score
Unit: No unit provided
Category: source-specific
Direction: Higher is better
Source type: vendor-reported
Evaluator: inceptionlabs.ai
Status: active
Retrieved at: 2026-10-04
Methodology: Vendor-reported result. Values read from the official image-only table. Quality composite and latency use Inception's voice-agent prompts/configuration.
Context: Source-specific observation; it is not a shared comparison cohort. The source table does not state a unit or scale for this numeric score.

Source inceptionlabs.ai

tau3 Airline (Show measurement, test conditions, and source)

Source value: 78
Score: 78
Metric: tau3 Airline reported score
Unit: No unit provided
Category: source-specific
Direction: Higher is better
Source type: vendor-reported
Evaluator: inceptionlabs.ai
Status: active
Retrieved at: 2026-10-04
Methodology: Vendor-reported result. Values read from the official image-only table. Quality composite and latency use Inception's voice-agent prompts/configuration.
Context: Source-specific observation; it is not a shared comparison cohort. The source table does not state a unit or scale for this numeric score.

Source inceptionlabs.ai

tau3 Retail (Show measurement, test conditions, and source)

Source value: 69.3
Score: 69.3
Metric: tau3 Retail reported score
Unit: No unit provided
Category: source-specific
Direction: Higher is better
Source type: vendor-reported
Evaluator: inceptionlabs.ai
Status: active
Retrieved at: 2026-10-04
Methodology: Vendor-reported result. Values read from the official image-only table. Quality composite and latency use Inception's voice-agent prompts/configuration.
Context: Source-specific observation; it is not a shared comparison cohort. The source table does not state a unit or scale for this numeric score.

Source inceptionlabs.ai

IFBench (Show measurement, test conditions, and source)

Source value: 67.4
Score: 67.4
Metric: IFBench reported score
Unit: No unit provided
Category: source-specific
Direction: Higher is better
Source type: vendor-reported
Evaluator: inceptionlabs.ai
Status: active
Retrieved at: 2026-10-04
Methodology: Vendor-reported result. Values read from the official image-only table. Quality composite and latency use Inception's voice-agent prompts/configuration.
Context: Source-specific observation; it is not a shared comparison cohort. The source table does not state a unit or scale for this numeric score.

Source inceptionlabs.ai

BFCL v4 multi-turn (Show measurement, test conditions, and source)

Source value: 60.5
Score: 60.5
Metric: BFCL v4 multi-turn reported score
Unit: No unit provided
Category: source-specific
Direction: Higher is better
Source type: vendor-reported
Evaluator: inceptionlabs.ai
Status: active
Retrieved at: 2026-10-04
Methodology: Vendor-reported result. Values read from the official image-only table. Quality composite and latency use Inception's voice-agent prompts/configuration.
Context: Source-specific observation; it is not a shared comparison cohort. The source table does not state a unit or scale for this numeric score.

Source inceptionlabs.ai

Sources and data date

Every statement links to its underlying documentation or leaderboard.

Type	Evidence and data date
Research status	Research date October 4, 2026. 1 source URLs checked. This documents the inspected sources, not an exhaustive inventory of every publication.
Additional source	Inception Mercury Voice announcement (retrieved October 3, 2026; October 4, 2026) · Inception Mercury Voice announcement · Editorial description reviewed October 4, 2026
Additional source	Inception model blog (retrieved October 3, 2026) · Inception model blog · Editorial description reviewed October 4, 2026