ctxdexSynthetic model data for learning
Active modelProprietaryAudio Generation2

Ctxdex Realtime 2

Agent MakersCtxdex Realtime

A second-generation realtime audio model in the Ctxdex Realtime family for speech-to-speech interaction, speech recognition and transcription, and text-to-speech synthesis.

AcceptsAudio, TextProducesAudio, TextReleasedAnnounced
View pricing

The pricing calculator is not yet built.

Data captured as of

Overview

Intended useBest suited for speech-to-speech interaction, speech recognition and transcription, text-to-speech synthesis, and tool-enabled workflows.
FamilyCtxdex RealtimeRealtime speech/audio interaction family.
AnnouncedAnnounced
DeprecatedNot applicableModel is current.
RetiredNot applicableModel is not retired.

What it does

Capabilities carry their support status; tasks are labels, because seven descriptions would bury the section.

Capabilities (1)

Tool Calling
Intrinsic support for requesting/invoking external tools/functions.

Tasks (3)

Speech RecognitionSpeech To SpeechText To Speech

Inputs & Outputs

Each modality is paired with the representation it is carried in, so a non-token model reads the same way a text one does.

Accepts

  • AudioSpeech Audio
  • TextChat Messages

Produces

  • AudioSpeech Audio
  • TextPlain Text

Technical Characteristics

A characteristic with no value states why it has none. An absence here is a recorded fact, not a gap in the page.

Maximum context tokens96000
Maximum native output tokens8000
Architecture classAUDIO_TRANSFORMER

Versions & Snapshots

Latest

Opening a snapshot shows what it resolves to. It does not switch this page into snapshot mode — the model's own values stay in place.

SnapshotReleasedStatusDetail
2026-03-03Active snapshot
2025-10-07Active snapshot

Availability

Current offering available

Agent Makers

First Party Managed Api

Active offering
Effective from

Serving surfaces

Audio Generation API · Active

Nimbrex Cloud

Cloud Managed Model

Active offering
Effective from

Serving surfaces

Audio Generation API · Active
Historical offeringsNone recorded
Downloadable distributionsNone recordedNo downloadable artefact has been published under these terms.

Serving Performance

Measured as of

Every figure is measured under the conditions beside it. Read without them, a latency number claims more than it can.

Measurement basis Synthetic ProfileMetrics 8
MetricTypical valueConditions
Audio Processing Realtime FactorProcessing time divided by input-audio duration; lower is faster.0.219Ratio
Capacity Mode: SHARED_ON_DEMANDDelivery Mode: SYNCHRONOUSProcessing Tier: STANDARDConcurrency Band: C1
Audio Processing Realtime FactorProcessing time divided by input-audio duration; lower is faster.0.251Ratio
Capacity Mode: SHARED_ON_DEMANDDelivery Mode: SYNCHRONOUSProcessing Tier: STANDARDConcurrency Band: C1
First Audio Response MsLatency to first generated audio response.185Ms
Capacity Mode: SHARED_ON_DEMANDDelivery Mode: REALTIMEInput Size Band: 5S_AUDIO_TURNProcessing Tier: STANDARDConcurrency Band: C1
First Audio Response MsLatency to first generated audio response.165Ms
Capacity Mode: SHARED_ON_DEMANDDelivery Mode: REALTIMEInput Size Band: 5S_AUDIO_TURNProcessing Tier: STANDARDConcurrency Band: C1
First Transcript Response MsLatency to first transcript/recognition result for streaming speech recognition.155Ms
Capacity Mode: SHARED_ON_DEMANDDelivery Mode: STREAMINGInput Size Band: 60S_AUDIOProcessing Tier: STANDARDConcurrency Band: C1
First Transcript Response MsLatency to first transcript/recognition result for streaming speech recognition.140Ms
Capacity Mode: SHARED_ON_DEMANDDelivery Mode: STREAMINGInput Size Band: 60S_AUDIOProcessing Tier: STANDARDConcurrency Band: C1
Turn Response Latency MsEnd-to-end conversational turn response latency under a defined turn workload.410Ms
Capacity Mode: SHARED_ON_DEMANDDelivery Mode: REALTIMEInput Size Band: 5S_AUDIO_TURNProcessing Tier: STANDARDConcurrency Band: C1
Turn Response Latency MsEnd-to-end conversational turn response latency under a defined turn workload.370Ms
Capacity Mode: SHARED_ON_DEMANDDelivery Mode: REALTIMEInput Size Band: 5S_AUDIO_TURNProcessing Tier: STANDARDConcurrency Band: C1

Pricing Reference

Reference offering

Agent Makers · First Party Managed Api

Effective fromCaptured as of
MeterBasisRate
Audio InputAudited audio-input minute rate$0.008 / 1Minute
Audio OutputAudited audio-output minute rate$0.055 / 1Minute
InputAudited text-input rate for realtime/speech context$0.6 / 1000000Token
View all pricing

Full pricing for a single offering is one step further in, from this model’s pricing page, because it is addressed by offering and this page carries the model. The calculator is disabled because it is not built.

1 other current offering is priced separately. This is the publisher's own offering, chosen as the reference — it is not the cheapest by definition. The meters below are on 2 different units and cannot be summed into a single rate. Rates are not comparable across offerings unless meter, unit and conditions match.

Derived chronology

Model Journey

Assembled from this model’s own lifecycle records, its snapshots, its providers’ offering and pricing history, and recorded succession. It is a view over those records, not a second source of truth, and undated lineage is deliberately not on it.

15 events2 scheduled ahead
  • Effective from
    PricingNormalUpcoming

    Agent Makers pricing scheduled

    Offering context: Agent Makers

    Scheduled to take effect on 2026-09-15. Not yet effective.

  • Effective from
    PricingNormalUpcoming

    Nimbrex Cloud pricing scheduled

    Offering context: Nimbrex Cloud

    Scheduled to take effect on 2026-09-15. Not yet effective.

  • Effective from
    SnapshotNormal

    Snapshot 2026-03-03 released

  • Effective from
    AvailabilityNormal

    Agent Makers · First Party Managed Api: Active

    Offering context: Agent Makers

    Synthetic commercial availability restored.

  • Effective from
    AvailabilityNormal

    Agent Makers · First Party Managed Api: Temporarily Unavailable

    Offering context: Agent Makers

    Synthetic temporary commercial withdrawal; not an operational incident.

  • Effective from
    PricingNormal

    Agent Makers pricing became effective

    Offering context: Agent Makers

    Version current, effective from 2026-01-01.

Open full journey →

The first control expands this section in place, without filters or navigation. The second opens the standalone journey, which adds filtering on three axes, year grouping, per-event provenance and a shareable address.

Every date is the source record’s own effective date — a model lifecycle date, a snapshot release, an offering’s effective date, a pricing version’s effective-from. None of them is a load or capture timestamp.

Licence & Use Conditions

LicenceAgent Makers Proprietary Model Terms
Version1.0
SPDX identifierNot applicableSPDX identifies open-source licences; these are proprietary terms.
Legal textNo public URI recordedThe terms exist and are held internally; no public address has been published.Captured as of

Use conditions (0)

None recordedNo condition constrains use of this model beyond its licence.

Lifecycle & Lineage

Direct lineage

Successor Of

Ctxdex Realtime 1

Recommended replacement for

Replacement for

Ctxdex Realtime 1

Effective from

Lifecycle history: Active on 2025-10-07.

Similar Industry Models

Captured as of

A positioning reference, not an equivalence claim. Each row states the axes it was drawn on, and covers only that part of this model's positioning.

GPT-Live-1

OpenAI

Provider documentation
CloseFull Duplex VoiceLow LatencyVoice Agent

Selected as a close current reference for Ctxdex Realtime 2 based on full duplex voice, low latency, voice agent.

Checked 2026-09-14 · High confidence

GPT-Realtime-2.1

OpenAI

Provider documentation
DirectRealtime AudioTool UseReasoning

Selected as a direct current reference for Ctxdex Realtime 2 based on realtime audio, tool use, reasoning.

Checked 2026-09-14 · High confidence

Gemini 3.1 Flash Live

Google

Provider documentation
CloseRealtime AudioFunction CallingMultimodal

Differences recorded: Preview.

Selected as a close current reference for Ctxdex Realtime 2 based on realtime audio, function calling, multimodal.

Checked 2026-09-14 · High confidence

Comparison method CTXDEX_COMPARABLES_V1.

Resolved from the ctxdex catalogue at build time. Every figure is a stored value or a derivation named in the read model; nothing on this page is computed in the browser.