Models
Every model CallMissed serves — Indic STT/TTS/LLM, fast direct-routed LLMs, first-party flagships, realtime voice and image — through one OpenAI-compatible API, plus 300+ more we deploy on demand.
Overview
158 models, one OpenAI-compatible API. Same auth, same request shape — change
the model field and nothing else.
| Group | What it is |
|---|---|
| Fast LLMs | Gemma 4 31B, the default for voice agents. |
| Indic models | STT, TTS and LLM built for 22 Indian languages. |
| Direct-routed LLMs | Sub-2s open-weights models: Kimi K2.5/K2.6/K2.7 Code and GPT-OSS. |
| First-party | gpt-6-*, gpt-6.1-sol, gpt-5.6-*, grok-4.3, DeepSeek-V4-*, realtime voice, plus first-party STT/TTS. Older GPT models are deprecated. |
| On demand | 300+ more we deploy on request. |
Models API
List all available models programmatically. No authentication required.
# List all models
curl https://api.callmissed.com/api/v1/models
# Filter by category: llm, stt, tts, image, embedding
curl https://api.callmissed.com/api/v1/models?category=llm
# Filter free-plan models only
curl https://api.callmissed.com/api/v1/models?free=true
# Get a specific model
curl https://api.callmissed.com/api/v1/models/sarvam-105b
# Which models each plan tier can call
curl https://api.callmissed.com/api/v1/models/accessEach entry includes id, name, description, category, owned_by, context_window, max_output_tokens (only where the model documents an output limit), pricing (US$), free, supports_streaming, supports_tools, supports_reasoning, supports_vision, zero_data_retention, processing ("india", "india_preferred" or "outside_india"; see Data residency), hosted_in_india (kept for compatibility; true only when processing is "india"), voice_capable (valid for POST /v1/voice/sessions), voice_agent_only, and speed_tier. A preview model also carries preview: true; the field is absent on generally available models. A model that is temporarily unavailable also carries status: "maintenance" and a status_message naming the alternative. Embedding models add dimensions; TTS models add their voice list. The response is {object: "list", total, data}.
The OpenAI-compatible listing at GET /v1/models (requires Authorization: Bearer cm_*) returns the same capability fields plus context_length (an alias of context_window for OpenAI-style clients), but a shorter list: it hides models that are only valid for voice sessions (gpt-realtime*, gpt-live-1, deepgram-voice-*). Use GET /api/v1/models for the full catalog. The Anthropic-shape listing at GET /anthropic/v1/models returns the same set inside Anthropic's {data, has_more, first_id, last_id} envelope.
Free Plan Models
The free tier includes 11 models across four categories. Use GET /api/v1/models?free=true to list them, or see the Model Access by Plan page for the full breakdown.
LLM (6 models)
| Model ID | Description |
|---|---|
sarvam-105b | 105B MoE — complex reasoning, Indic languages |
sarvam-105b-conversations | 105B MoE tuned for conversation and voice — 32K context, tool calling |
kimi-k2.5 | Moonshot K2.5 — 262K context, reasoning, vision |
kimi-k2.6 | Moonshot K2.6 — improved reasoning + coding, 262K context |
kimi-k2.7-code | Moonshot K2.7 Code — frontier 1T-param agentic coding, 262K context, vision + tools |
gpt-oss-120b | GPT-OSS 120B — open-weights large model |
STT (1 model)
| Model ID | Description |
|---|---|
saaras:v4 | 23 langs (22 Indic + English) — five output modes: transcribe, translate, verbatim, transliterate, code-mix |
TTS (2 models)
| Model ID | Description |
|---|---|
bulbul:v3 | 37 voices, 11 Indian languages |
bulbul:v4-flash | 222 persona voices, 9 Indian languages + Hinglish, low latency |
Embedding (2 models)
| Model ID | Description |
|---|---|
text-embedding-3-small | 1536 dimensions, 8,192-token inputs — best price/performance |
text-embedding-3-large | 3072 dimensions, 8,192-token inputs — highest accuracy |
A free-plan key calling a paid model gets 403 model_not_available — it is not billed, it is refused. Upgrade to Starter or above first.
The Claude models (claude-opus-5-5, claude-sonnet-5-5, claude-haiku-5-5) are the exception: they need the Pro plan or higher. A free or Starter key gets 403 model_not_available with the message Model '<id>' requires the Pro plan or higher. Upgrade at https://console.callmissed.com/org/plans.
All other models — including kimi-k3.1-pro, glm-5.3, glm-5.3-flash, deepseek-v4.1-flash, deepseek-v4.1-flash-1017, gemma-4-31b, deepgram-think-fast, deepgram-phonellm-mini, the gemini-* chat models, first-party IDs (gpt-4o, gpt-4.1, gpt-5-mini, gpt-5.5, gpt-5.6-*, gpt-6-*, gpt-6.1-sol, grok-4.3, DeepSeek-V4-*, gpt-realtime*, gpt-live-1, first-party STT/TTS), the other speech models (gnani-*, ink-*, sonic-3.6, assemblyai-*, inworld-*, rime-coda, fish-audio-*, gradium-*, deepgram-flux-tts, microsoft-*), the Deepgram direct line (deepgram-nova-3, deepgram-flux-general-en/multi, deepgram-nova-2*, deepgram-enhanced*, deepgram-base*, deepgram-whisper-*, deepgram-aura-2, deepgram-aura-1, Deepgram Voice Agent deepgram-voice-* ids, the deepgram-summarize/topics/sentiment/intents Audio Intelligence features, and the deepgram-text-summarize/topics/sentiment/intents Text Intelligence features), and paid image models (flux-2-pro, flux-1.1-pro, gpt-image-2.5-*, gpt-image-2, gpt-image-1.5, nano-banana-*, gemini-3.1-flash-lite-image) — require a paid plan (Starter or higher), except Claude models, which need Pro or higher.
Pricing
All models are pay-per-use. Pricing is in US$ at US$1 = ₹96; you pay in credits, where 1 credit = ₹1 ≈ US$0.0104.
| Model | Input / 1M tokens | Output / 1M tokens |
|---|---|---|
sarvam-105b | $0.1563 (₹15) | $0.625 (₹60) |
sarvam-105b-conversations | $0.1563 (₹15) | $0.625 (₹60) |
gpt-5.6-sol | $4.167 | $20.83 |
gpt-5.6-terra | $2.083 | $12.5 |
gpt-5.6-luna | $0.2083 | $1.25 |
gpt-6-sol | $2.083 | $10.42 |
gpt-6-luna | $0.1042 | $0.5208 |
gpt-6.1-sol | $2.083 | $10.42 |
claude-opus-5-5 (Pro plan and up) | $4.167 | $20.83 |
claude-sonnet-5-5 (Pro plan and up) | $2.083 | $10.42 |
claude-haiku-5-5 (Pro plan and up) | $0.1042 ($0.5208 for prompts over 100K tokens) | $0.5208 ($2.604 for prompts over 100K tokens) |
gpt-realtime | $4.167 | $16.67 |
gpt-realtime-mini | $0.625 | $2.5 |
gpt-realtime-2 | $4.167 | $25 |
gpt-realtime-1.5 | $4.167 | $16.67 |
gpt-realtime-2.1 | $4.167 | $25 |
gpt-realtime-2.1-mini | $0.625 | $2.5 |
gpt-live-1 | per-minute only | $0.05208/min (voice-agent only) |
deepgram-voice-* | per-minute Voice Agent tier | Standard $0.07813/min, Advanced $0.1698/min (voice-agent only) |
| STT Model | Price |
|---|---|
saaras:v4 | $0.3125 / hour (₹30/hr) |
gnani-prisma-v3.0 | $0.2813 / hour |
ink-whisper | $0.1875 / hour |
ink-2 | $0.5625 / hour (voice sessions only) |
assemblyai-universal-3.6-pro | $0.4688 / hour (voice sessions only) |
inworld-stt-1 | $0.1563 / hour (voice sessions only) |
microsoft-speech-stt | $1.0417 / hour (voice sessions only) |
mai-transcribe-2 | $0.375 / hour (preview) |
scribe_v2 / scribe_v2_medical | $0.2292 / hour |
scribe_v2_realtime | $0.4063 / hour (voice sessions only) |
deepgram-nova-3 | $0.325 / hour (files); live voice sessions $0.4792 / hour, or $0.5729 / hour with multi |
deepgram-nova-3-medical | $0.325 / hour (files); live voice sessions $0.4792 / hour, or $0.5729 / hour with multi |
deepgram-nova-3-pharma | $0.325 / hour (files), $0.4792 / hour (live voice sessions) |
deepgram-flux-general-en | $0.4792 / hour |
deepgram-flux-general-multi | $0.4896 / hour |
deepgram-nova-2 (+ domain variants) | $0.3646 / hour |
deepgram-nova / deepgram-whisper-* | $0.3646 / hour |
deepgram-enhanced (+ variants) | $1.031 / hour |
deepgram-base (+ variants) | $0.9063 / hour |
| TTS Model | Price |
|---|---|
bulbul:v3 | $0.3125 / 10K chars (₹30/10K) |
bulbul:v4-flash | $0.2083 / 10K chars (₹20/10K) |
gnani-timbre-v3.0 | $0.2813 / 10K chars |
deepgram-aura-2 | $0.3125 / 10K chars |
deepgram-aura-1 | $0.1563 / 10K chars |
sonic-3.6 | $0.5208 / 10K chars |
deepgram-flux-tts | $0.4688 / 10K chars |
rime-coda | $0.5208 / 10K chars (voice sessions only) |
inworld-tts-2 | $0.2604 / 10K chars (voice sessions only) |
inworld-tts-2-flash | $0.1563 / 10K chars (voice sessions only) |
fish-audio-drama-3 | $0.5208 / 10K chars (voice sessions only) |
gradium-tts-beta | $0.50 / 10K chars (voice sessions only) |
microsoft-neural-tts | $0.1563 / 10K chars (voice sessions only) |
mai-voice-2.1 | $0.2292 / 10K chars (preview) |
mai-voice-2.1-flash | $0.1563 / 10K chars (preview) |
eleven_v4 / eleven_v3 / eleven_multilingual_v2 | $0.8333 / 10K chars |
eleven_v4_turbo / eleven_v3_conversational / eleven_flash_v2_5 / eleven_turbo_v2_5 / eleven_flash_v2 / eleven_turbo_v2 | $0.4167 / 10K chars |
| Intelligence feature (not a model ID) | Price |
|---|---|
deepgram-summarize / -topics / -sentiment / -intents (audio) | $0.0003125 / 1K input + $0.000625 / 1K output tokens |
deepgram-text-summarize / -text-topics / -text-sentiment / -text-intents | $0.0003125 / 1K input + $0.000625 / 1K output tokens |
These eight values go in the features field, not in model. They are not
catalog models — GET /api/v1/models/deepgram-summarize returns 404.
Full pricing for all models is available via the API: GET /api/v1/models
import requests
# List all LLM models
models = requests.get("https://api.callmissed.com/api/v1/models?category=llm").json()
for m in models["data"]:
print(f"{m['id']} — {m['name']} ({m['context_window']} tokens) {'FREE' if m['free'] else 'PAID'}")Indic Models
Speech to Text
| Model | Description | Languages |
|---|---|---|
saaras:v4 | Five output modes on one model — transcribe, translate, verbatim, transliterate, code-mix | 23 languages (22 Indic + English) |
gnani-prisma-v3.0 | Gnani's newest India-first telephony STT — automatic language detection (indic-auto), mixed-language speech, keyterm boosting | English + 9 Indian languages (10 total) |
For 99-language general-purpose transcription and translation, see whisper. For diarization + smart-format on calls, see deepgram-nova-3. Both live under the audio model routes.
ink-whisper also covers Hindi, Urdu and Tamil as part of its 100-language set at $0.1875 / hr — cheaper than the Indic-specialist models, though without their code-mix output modes. ink-2 covers Hindi among its five languages (en, fr, hi, ja, es) but no other Indic language. microsoft-speech-stt (voice sessions only, processed in India) hears 81 languages, including Hindi, Bengali, Tamil, Telugu, Marathi, Gujarati, Kannada, Malayalam, Punjabi, Odia, Assamese, Urdu, Bhojpuri and Indian English, but only one language per call. assemblyai-universal-3.6-pro (voice sessions only) hears Hindi, Marathi and Urdu, plus Indian English (en-IN), among its 32 languages.
Text to Speech
| Model | Description | Voices |
|---|---|---|
bulbul:v3 | Natural TTS — 37 voices, 11 Indian languages | shubh (default) + 36 more |
bulbul:v4-flash | Low-latency TTS — 9 Indian languages + Hinglish; no Malayalam or Odia (use bulbul:v3) | 222 persona voices, shubh_enhi_ads (default) |
gnani-timbre-v3.0 | Gnani Timbre v3 — India-first neural TTS, streaming; every voice speaks all 10 Indian languages plus Hinglish; supports expressive audio tags in the text | 12 voices (Shivani default) |
sonic-3.6 | Cartesia Sonic 3.6 — most natural conversational speech, 44 languages with native-quality Hindi | Searchable public library + 16 featured aliases (skylar default) |
rime-coda | Rime Coda — voice sessions only; 8 languages including Hindi | 13 voices, two native Hindi (nadi default) |
inworld-tts-2 / inworld-tts-2-flash | Inworld TTS-2 — voice sessions only; 94 languages including Hindi, Marathi, Bengali, Tamil, Telugu, Gujarati, Kannada, Malayalam, Punjabi, Urdu and Odia | 14 voices (riya default) |
microsoft-neural-tts | Microsoft Neural voices — voice sessions only; processed in India; 13 languages including Hindi, Bengali, Tamil, Telugu, Marathi, Gujarati, Kannada, Malayalam, Punjabi, Urdu, Odia and Assamese | 57 voices (en-in-neerja default) |
For low-latency English / Spanish voice agents, see deepgram-aura-2. For steerable multilingual delivery, see gpt-4o-mini-tts.
Chat Completion (LLM)
| Model | Params | Context | Best For |
|---|---|---|---|
sarvam-105b | 105B MoE | 128K tokens | Complex reasoning, agentic tasks, long documents |
sarvam-105b-conversations | 105B MoE | 32K tokens | Conversation and voice agents, tool calling |
glm-5.3 | Z.ai GLM 5.3 | 1M tokens | Long-context reasoning, tool calling (text only) |
deepseek-v4.1-flash | DeepSeek V4.1 Flash | 1M tokens | Long-document analysis, agentic workflows, tool calling (text only) |
glm-5.3-flash | Z.ai GLM 5.3 Flash | 1M tokens | Fast, low-cost coding agents and tool use, with image input |
kimi-k3.1-pro | Moonshot Kimi K3.1 Pro (2.8T MoE, 104B active) | 1M tokens | Agentic coding and research, long-horizon tool use, image input, JSON mode |
deepseek-v4.1-flash-1017 | DeepSeek V4.1 Flash 1017 | 1M tokens | Long-document analysis, agentic workflows, tool calling, image input |
gemma-4-31b | Google Gemma 4 31B | 128K tokens | Fast answers with thinking off, image input, tool calling, voice agents |
deepgram-think-fast | Deepgram Think-Fast (early access) | 32K tokens | Voice agents: follows a system prompt, calls tools, writes replies meant to be spoken (text only) |
deepgram-phonellm-mini | phoneLLM mini by Pipecat (early access) | 32K tokens | Voice pipelines where speed matters most, tool calling (text only) |
glm-5.3 and deepseek-v4.1-flash always reason. reasoning_effort takes "low", "high" or "max" (the default when omitted); "none" and "minimal" map to "low", "medium" to "high" and "xhigh" to "max". Reasoning tokens bill as output. glm-5.3-flash always reasons too, with levels "low", "high" and "max" and "high" as the default when omitted; its "none", "minimal", "medium" and "xhigh" mappings are the same as glm-5.3. deepseek-v4.1-flash-1017 is the October 17 snapshot of DeepSeek V4.1 Flash: it reasons at "high" by default and "none" switches thinking off. Both new models accept image input as base64 data URIs (we fetch and inline https:// image URLs for you), tool calling and streaming; glm-5.3-flash also supports structured output with a JSON schema, and deepseek-v4.1-flash-1017 supports tool_choice: "required". When you omit max_tokens we send 16,384. gemma-4-31b answers without thinking by default. reasoning_effort: "low", "medium" or "high" switch thinking on; the level does not reliably change how much it thinks. "none" and "minimal" keep it off. Both are paid-plan models.
Both sarvam-105b models support hybrid thinking via reasoning_effort: "low" | "high" | "max", and "xhigh" maps to "max". "none" and "minimal" switch thinking fully off. "medium" is not one of their levels, so it is dropped and the model runs at its default. Other models with a full off switch: kimi-k2.5, kimi-k2.6, gemma-4-31b, DeepSeek-V4-Pro, DeepSeek-V4-Flash, or a GPT-5.5 / GPT-5.6 / GPT-6 model with reasoning_effort: "none". See the per-model matrix.
Audio Models
See the Pricing page for current rates.
Speech to Text
| Model | Languages | Best for | Price |
|---|---|---|---|
whisper | 99 with auto-detect | Whisper batch + translate | $0.4167 / hour |
gpt-4o-transcribe | Streaming | Higher-accuracy OpenAI transcription | $0.4167 / hour |
gpt-4o-mini-transcribe | Streaming | Low-cost OpenAI transcription | $0.25 / hour |
gpt-4o-transcribe-diarize | File transcription with speaker labels (not available for voice sessions) | Multi-speaker meetings / calls | $0.4167 / hour |
mai-transcribe-2 Preview | 60 with auto-detect, Hinglish code-switching (file transcription only) | Microsoft AI file transcription, up to 2 hours per file | $0.375 / hour |
scribe_v2 | 100 with auto-detect, 13 Indian languages | ElevenLabs' highest-accuracy file transcription | $0.2292 / hour |
scribe_v2_medical | 100 with auto-detect | Clinical audio (file transcription) | $0.2292 / hour |
scribe_v2_realtime | 100 with auto-detect (voice sessions only) | Real-time voice agents | $0.4063 / hour |
Deepgram (direct) — the full Deepgram speech-to-text line, billed per audio hour at the rates below:
| Model | Languages | Best for | Price |
|---|---|---|---|
deepgram-flux-general-en | English | Conversational voice agents — model-native turn detection, ultra-low latency | $0.4792 / hour |
deepgram-flux-general-multi | 10 (multilingual) | Multilingual voice agents with code-switching | $0.4896 / hour |
deepgram-nova-3 | 45+ incl. multi | Flagship general-purpose ASR, keyterm prompting, PII redaction | $0.325 / hour (files); live voice sessions $0.4792 / hour, or $0.5729 / hour with multi |
deepgram-nova-3-medical | English, or multi (10 languages) | Clinical / medical terminology | $0.325 / hour (files); live voice sessions $0.4792 / hour, or $0.5729 / hour with multi |
deepgram-nova-3-pharma | English | Drug names and pharmaceutical terminology, keyterm prompting | $0.325 / hour (files), $0.4792 / hour (live voice sessions) |
deepgram-nova-2 | 36 incl. multi | High-accuracy ASR + filler-word detection | $0.3646 / hour |
deepgram-nova-2-{meeting,phonecall,finance,conversationalai,voicemail,video,medical,drivethru,automotive,atc} | English | Domain-tuned Nova-2 variants | $0.3646 / hour |
deepgram-nova / -phonecall / -medical | en/es/hi | Legacy Nova-1 | $0.3646 / hour |
deepgram-enhanced (+ meeting/phonecall/finance) | 13 | Legacy, keyword boosting | $1.031 / hour |
deepgram-base (+ 6 variants) | 17 | Legacy, high-volume batch | $0.9063 / hour |
deepgram-whisper-{tiny,base,small,medium,large} | 99 | Whisper in five sizes, batch transcription | $0.3646 / hour |
Text to Speech
| Model | Languages | Voices | Price |
|---|---|---|---|
deepgram-aura-2 | en/es/de/fr/nl/it/ja | 90+ (thalia default) | $0.3125 / 10K chars |
deepgram-aura-1 | English | 12 (asteria default) | $0.1563 / 10K chars |
deepgram-flux-tts | 42 (149 GA voices in 10 languages, 117 Early Access voices in 32) | 266 (priya default) | $0.4688 / 10K chars |
gpt-4o-mini-tts | Multilingual steerable | 6 OpenAI voices | $0.2083 / 10K chars |
mai-voice-2.1 Preview | 23 languages | 97 (en-in-priya default; hi-in-arjun for Hindi) | $0.2292 / 10K chars |
mai-voice-2.1-flash Preview | 23 languages | 97, low latency for live calls | $0.1563 / 10K chars |
eleven_v4 | 91 incl. 12 Indian languages | 21 premade + any ElevenLabs voice ID (sarah default) | $0.8333 / 10K chars |
eleven_v4_turbo | 91 incl. 12 Indian languages | same, tuned for live calls | $0.4167 / 10K chars |
eleven_v3 | 85 incl. 11 Indian languages | same, dramatic delivery | $0.8333 / 10K chars |
eleven_v3_conversational | 85 incl. 11 Indian languages | same, conversational agents | $0.4167 / 10K chars |
eleven_multilingual_v2 | 29 incl. Hindi, Tamil | same, long-form narration | $0.8333 / 10K chars |
eleven_flash_v2_5 | 32 incl. Hindi, Tamil | same, fastest multilingual | $0.4167 / 10K chars |
eleven_turbo_v2_5 | 32 incl. Hindi, Tamil | same (equivalent to Flash v2.5) | $0.4167 / 10K chars |
eleven_flash_v2 | English | same, fastest English | $0.4167 / 10K chars |
eleven_turbo_v2 | English | same (equivalent to Flash v2) | $0.4167 / 10K chars |
The mai-* models are preview: Microsoft ships them without an SLA and may change or retire them at short notice. If one is retired, the ID keeps working and is answered by a named successor, as in Retired and deprecated models, so an agent that uses it keeps working but may sound different. Use a generally available model for production-critical agents. Preview models are paid-plan only, are never a default, and are not processed in India.
Aura 2 returns linear16 PCM streamed at 24 kHz for low-latency playback. Output formats may vary as models are updated.
The nine eleven_* models are ElevenLabs' full current text-to-speech line, on POST /v1/audio/speech and in voice agents and voice sessions; scribe_v2 and scribe_v2_medical transcribe files and scribe_v2_realtime serves voice sessions. See Text to Speech, Speech to Text and Voices.
Deepgram Flux TTS (deepgram-flux-tts) is served on this endpoint, in voice agents and voice sessions, and in the managed Voice Agent (selectable with tts_engine: "flux", billed inside the per-minute voice rate). It has 266 voices in 42 languages. Without stream, the input is limited to 400 characters; set "stream": true with pcm, wav, mulaw or alaw for up to 2000. See Text to Speech and Voices.
Voice-session-only models
These models run inside voice agents and voice sessions only: set them as stt_model or tts_model. POST /v1/audio/transcriptions and POST /v1/audio/speech return 400 for them. Voices, languages and markup support are on the Voices and Speech to Text pages.
| Model | Type | Languages | Price |
|---|---|---|---|
assemblyai-universal-3.6-pro | STT, built-in turn detection | 50, including Hindi, Marathi, Urdu and Indian English | $0.4688 / hour |
inworld-stt-1 | STT | English only | $0.1563 / hour |
microsoft-speech-stt | STT, processed in India | 81 (148 locales), including 13 Indian languages, one per call | $1.0417 / hour |
scribe_v2_realtime | STT, auto-detects the language | 100, including 13 Indian languages | $0.4063 / hour |
rime-coda | TTS, 13 voices | 8, including Hindi | $0.5208 / 10K chars |
inworld-tts-2 | TTS, 14 voices | 94, including 11 Indian languages | $0.2604 / 10K chars |
inworld-tts-2-flash | TTS, 14 voices, faster | 94, including 11 Indian languages | $0.1563 / 10K chars |
fish-audio-drama-3 | TTS, 7 voices | 13, no Indian languages | $0.5208 / 10K chars |
gradium-tts-beta | TTS, 13 voices | 5, no Indian languages | $0.50 / 10K chars |
microsoft-neural-tts | TTS, 57 voices, processed in India | 13, including 12 Indian languages | $0.1563 / 10K chars |
Deepgram managed voice-agent models
The deepgram-voice-* models run the whole conversation (language model, speech recognition and voice) inside Deepgram's managed Voice Agent and bill one per-minute rate: Standard $0.07813/min, Advanced $0.1698/min. They are valid on voice agents and voice sessions only, not on POST /v1/chat/completions. Four were added on 2026-10-30:
| Model | Tier | Notes |
|---|---|---|
deepgram-voice-anthropic-claude-sonnet-5-5 | Advanced | Strongest reasoning of the four |
deepgram-voice-anthropic-claude-haiku-5-5 | Standard | Fast, low cost |
deepgram-voice-deepgram-think-fast | Standard | Early access. Conversations on this model are used by the provider to improve its models (the account's usual data-sharing opt-out does not apply to it) |
deepgram-voice-deepgram-phonellm-mini-alpha-2-fp8 | Standard | Early access |
Deepgram Think-Fast and phoneLLM mini (early access)
deepgram-think-fast and deepgram-phonellm-mini are chat models for voice agents, called on POST /v1/chat/completions like any other model and selectable as the llm_model of a voice agent (speech-to-text, language model, text-to-speech). They are not the managed deepgram-voice-* bundle above, which bills one per-minute rate and has no chat endpoint. Both are $0.3646 per 1M input tokens and $0.9375 per 1M output tokens, with no cached-input price, and are paid-plan only.
deepgram-think-fast | deepgram-phonellm-mini | |
|---|---|---|
| Maker | Deepgram | Pipecat |
| Context window | 32,768 tokens | 32,768 tokens |
| Tool calling, streaming | Yes | Yes |
| Vision | No (text only) | No (text only) |
| Reasoning | Yes, via reasoning_effort (none is the default); trace in reasoning_content | No |
| Training on your data | Conversations on this model are used by the provider to improve its models; the opt-out does not apply to it. | Excluded from the provider's model training |
deepgram-think-fast follows a system prompt, calls tools, and writes replies meant to be spoken: it often says a short line such as "Sure, booking that now" alongside a tool call. Keep reasoning off for voice. In our tests "minimal" to "high" took 2 to 5 seconds to answer, against about 0.3 seconds with "none". deepgram-phonellm-mini is an open-weights model based on NVIDIA Nemotron 3 Nano (30B mixture-of-experts, 3.5B active parameters), trained on synthetic voice-agent sessions; use temperature 0.0, which voice agents use by default. Measured from our India servers, a one-sentence reply takes 0.3 to 0.8 seconds on deepgram-think-fast and 0.3 to 0.4 seconds on deepgram-phonellm-mini. Languages, from our own tests (three receptionist questions per language): deepgram-think-fast answered correctly in English, Hindi, Marathi, Telugu and Kannada, stayed fluent but got a day, time or price wrong in Tamil, Gujarati and Bengali, and got the days and fee wrong in Malayalam; deepgram-phonellm-mini answered correctly only in English and Hindi (Hindi terse) and mixed scripts or invented facts in every other Indian language tried. Validating a voice agent's config flags a language outside those lists: an error for deepgram-phonellm-mini, a warning for deepgram-think-fast, and the same for language: "multi".
Early access: the provider says request fields and limits may change, and there is no SLA. Limits on these two models:
- Processed outside India. An India-only key or account gets
400 region_unavailable, and a voice agent withhosting: "india"cannot select them. See Data residency. - Accepted fields:
model,messages,temperature(0 to 2),reasoning_effort,tools(function tools),tool_choice("auto"or"none"only; any other value returns400 unsupported_parameter),streamandstream_options. response_format(JSON mode, structured outputs) returns400 unsupported_parameter.- Ignored, not sent to the model:
max_tokens,top_p,stop,seed,frequency_penalty,presence_penalty,logprobs,nandparallel_tool_calls. - Image and file content parts are not supported.
Audio Intelligence (Deepgram)
Deepgram Audio Intelligence runs analysis over an uploaded audio file via POST /v1/audio/intelligence (English only, 150K input-token limit, 25 MB upload limit). Billed at $0.0003125/1K input + $0.000625/1K output tokens, plus the transcription the analysis runs on at the per-hour rate of the matching deepgram-* STT model (e.g. model=nova-3 bills at the deepgram-nova-3 rate, $0.3021 / hour).
| Feature | Model ID | Returns |
|---|---|---|
| Summarization | deepgram-summarize | A concise summary of the audio |
| Topic Detection | deepgram-topics | Per-segment topics with confidence |
| Sentiment Analysis | deepgram-sentiment | Per-segment + average sentiments |
| Intent Recognition | deepgram-intents | Per-segment intents with confidence |
Request multiple features in one call with a comma-separated features form field (e.g. features=deepgram-summarize,deepgram-sentiment).
Text Intelligence (Deepgram)
Deepgram Text Intelligence runs the same four analyses over text input (a string or a hosted text URL) via POST /v1/text/intelligence (English only, 150K input-token limit). Token-billed at $0.0003125/1K input + $0.000625/1K output. Requires the llm key permission.
| Feature | Model ID | Returns |
|---|---|---|
| Summarization | deepgram-text-summarize | A concise summary of the text |
| Topic Detection | deepgram-text-topics | Per-segment topics with confidence |
| Sentiment Analysis | deepgram-text-sentiment | Per-segment + average sentiments |
| Intent Recognition | deepgram-text-intents | Per-segment intents with confidence |
Send a JSON body with features (array or comma-separated string) and exactly one of text or url:
{
"features": ["deepgram-text-summarize", "deepgram-text-sentiment"],
"text": "Your text to analyze here."
}Direct-Routed LLMs
Low-latency models routed directly through CallMissed — sub-2s end-to-end on small prompts and free-tier eligible per the reasoning_effort matrix.
| Model ID | Creator | Context |
|---|---|---|
kimi-k2.5 | Moonshot AI | 262K |
kimi-k2.6 | Moonshot AI | 262K |
kimi-k2.7-code | Moonshot AI | 262K |
gpt-oss-120b | OpenAI (open-weights) | 131K |
Retired and Deprecated Models
Retired on 2026-10-17, 2026-10-30 and 2026-11-11. These IDs still work. Each one is answered by the
model in the second column, and the response model field keeps the ID you
sent. They are no longer listed by GET /api/v1/models. Requests are billed
at the price shown.
| Retired ID | Answered by | Billed at |
|---|---|---|
glm-4.7-flash | gpt-oss-120b | Unchanged ($0.5208 in / $2.083 out per 1M) |
glm-5.2 | kimi-k2.7-code | Unchanged ($1.969 in / $6.188 out per 1M) |
nemotron-3-super | kimi-k2.6 | Unchanged ($1.563 in / $6.25 out per 1M) |
gemma-4-26b-a4b-it | kimi-k2.5 | kimi-k2.5 price ($0.8438 in / $4.219 out per 1M) |
mistral-small-3.1 | kimi-k2.5 | kimi-k2.5 price ($0.8438 in / $4.219 out per 1M) |
flux-2-klein-9b, phoenix-1.0 | flux-2-pro | Unchanged ($0.1042 / image) |
lucid-origin | flux-2-pro | Unchanged ($0.08333 / image) |
sdxl-lightning, dreamshaper-8-lcm | flux-2-pro | flux-2-pro price ($0.1042 / image) |
aura-2-en, aura-2-es | deepgram-aura-2 (same voices) | Unchanged ($0.4167 / 10K chars) |
melotts | gpt-4o-mini-tts | gpt-4o-mini-tts price ($0.2083 / 10K chars) |
nova-3 | deepgram-nova-3 | Unchanged ($0.5208 / hour) |
gnani-prisma-v2.5 (retired 2026-10-30) | gnani-prisma-v3.0 | Keeps its price ($0.2813 / hour) |
whisper-large-v3-turbo | whisper (also translates) | whisper price ($0.4167 / hour) |
kimi-k2.5-fast (retired 2026-11-11) | kimi-k2.5 | kimi-k2.5 price ($0.8438 in / $4.219 out per 1M) |
The retired LLM, speech-to-text and text-to-speech IDs stay on the free plan,
except kimi-k2.5-fast, which was never on it and stays paid-only.
The retired image IDs do not: the free plan has no image model, and a free key
gets 403 model_not_available.
Deprecated, sunset 2026-11-16. These models are still listed and served as
themselves until the sunset date. The catalogue marks them with
deprecated: true, sunset_date and successor, and every API response to a
request that names one carries Deprecation, Sunset and
Link: <…/changelog>; rel="deprecation" headers. From 2026-11-16 a request
that names one is answered by, and billed as, its successor.
| Deprecated ID | Successor |
|---|---|
gpt-4o, gpt-4.1, gpt-5.5, gpt-5.6-sol, gpt-5.6-terra | gpt-6-sol |
gpt-5-mini, gpt-5.6-luna | gpt-6-luna |
gemini-3.5-flash, gemini-3.6-flash, gemini-3.7-flash | gemini-3.8-flash |
gemini-3.1-flash-lite | gemini-3.5-flash-lite |
gpt-realtime, gpt-realtime-1.5, gpt-realtime-2 | gpt-realtime-2.1 |
gpt-realtime-mini | gpt-realtime-2.1-mini |
gpt-image-1.5, gpt-image-2 | gpt-image-2.5-flare |
flux-1.1-pro | flux-2-pro |
deepgram-nova, deepgram-nova-2, deepgram-enhanced, deepgram-base | deepgram-nova-3 |
deepgram-nova-medical, deepgram-nova-2-medical | deepgram-nova-3-medical |
deepgram-nova-phonecall, deepgram-enhanced-phonecall, deepgram-base-phonecall | deepgram-nova-2-phonecall |
deepgram-enhanced-meeting, deepgram-base-meeting | deepgram-nova-2-meeting |
deepgram-enhanced-finance, deepgram-base-finance | deepgram-nova-2-finance |
deepgram-base-conversationalai, deepgram-base-voicemail, deepgram-base-video | the matching deepgram-nova-2-* model |
Where a successor costs more (for example gpt-4.1 output, gemini-3.1-flash-lite), the request bills at the successor's price from the sunset date; switch earlier to test.
Models on Demand
GET /api/v1/models lists everything in the catalogue today: 147 model IDs.
Models marked status: "maintenance" return 503 until they are back.
Beyond that we deploy 300+ further models on demand on CallMissed
infrastructure. Send the model you need and your expected throughput to
[email protected]. Once deployed it appears in your GET /api/v1/models
response with a plain CallMissed ID and published per-token pricing, on the
same /v1/chat/completions endpoint as every other model. Same key, same
credit balance, no new SDK.
First-Party Models
Credit-covered first-party models. Use the bare model ID in API requests — e.g. gpt-4o.
| Model ID | Type | Notes |
|---|---|---|
gpt-4o | LLM | Multimodal text + vision, 128K context — deprecated, sunset 2026-11-16 → gpt-6-sol |
gpt-4.1 | LLM | Long-context (300K) multimodal — deprecated, sunset 2026-11-16 → gpt-6-sol |
gpt-5-mini | LLM | Fast reasoning, 400K context — deprecated, sunset 2026-11-16 → gpt-6-luna |
gpt-5.5 | LLM | GPT-5.5 reasoning flagship, 1.05M context, vision + tools — deprecated, sunset 2026-11-16 → gpt-6-sol |
gpt-5.6-sol | LLM | GPT-5.6 flagship, 1.05M context, vision + tools — deprecated, sunset 2026-11-16 → gpt-6-sol |
gpt-5.6-terra | LLM | GPT-5.6 balanced intelligence/cost, 1.05M context — deprecated, sunset 2026-11-16 → gpt-6-sol |
gpt-5.6-luna | LLM | GPT-5.6 fast + affordable, 1.05M context — deprecated, sunset 2026-11-16 → gpt-6-luna |
gpt-6-sol | LLM | GPT-6 frontier reasoning, 1.05M context, vision + tools |
gpt-6-luna | LLM | GPT-6 efficient, high-volume, 1.05M context |
gpt-6.1-sol | LLM | GPT-6.1 near-frontier for complex coding and professional work, 1.05M context, vision + tools |
claude-opus-5-5 | LLM | Anthropic Claude Opus 5.5 for long-running agentic coding and knowledge work, 1M context, vision + tools (Pro plan and up) |
claude-sonnet-5-5 | LLM | Anthropic Claude Sonnet 5.5, the best combination of speed and intelligence, 1M context, vision + tools (Pro plan and up) |
claude-haiku-5-5 | LLM | Anthropic Claude Haiku 5.5 for high-volume, latency-sensitive work, 1M context, vision + tools (Pro plan and up) |
grok-4.3 | LLM | xAI Grok, 200K context |
kimi-k3.1-pro | LLM | Moonshot K3.1 Pro, 1M context, vision + tools + reasoning (paid plans only) |
DeepSeek-V4-Pro | LLM | Flagship DeepSeek reasoning, 1M context, tools (text-only) |
DeepSeek-V4-Flash | LLM | Fast DeepSeek reasoning, 1M context, tools |
gpt-realtime | Realtime voice | OpenAI flagship speech-to-speech (10 concurrent), live — deprecated, sunset 2026-11-16 → gpt-realtime-2.1 |
gpt-realtime-mini | Realtime voice | Lowest-cost realtime, ~3× cheaper than gpt-realtime, live — deprecated, sunset 2026-11-16 → gpt-realtime-2.1-mini |
gpt-realtime-2 | Realtime voice | Answered by gpt-realtime-2.1 — deprecated, sunset 2026-11-16 → gpt-realtime-2.1 |
gpt-realtime-1.5 | Realtime voice | Pinned 1.5 snapshot of gpt-realtime, live — deprecated, sunset 2026-11-16 → gpt-realtime-2.1 |
gpt-realtime-2.1 | Realtime voice | Latest realtime — better recognition, silence/interrupt handling, live |
gpt-realtime-2.1-mini | Realtime voice | Distilled low-cost 2.1 realtime, live |
gpt-live-1 | Realtime voice | GPT Live speech-to-speech — audio + text only, 14 voices, $0.05208/min, live |
whisper | STT | OpenAI Whisper — 99 langs |
gpt-4o-transcribe | STT | Streaming transcription |
gpt-4o-mini-transcribe | STT | Low-cost streaming STT |
gpt-4o-transcribe-diarize | STT | Speaker diarization |
gpt-4o-mini-tts | TTS | Steerable OpenAI TTS, 13 voices |
See Credits & Rate Limits for per-model USD pricing.
Full Model Catalog
A curated, representative slice of the 158 models (72 LLM · 50 STT · 25 TTS · 9 image · 2 embedding) served by GET /api/v1/models as of the latest deploy — the per-domain variants of the direct Deepgram STT line and the deepgram-voice-* managed LLM ids are covered in their own sections above rather than repeated below. For live pricing and capability flags (supports_vision, supports_tools, free), query the API — it always reflects the current catalog.
LLM (44 models)
| Model ID | Description | Context | Free | Pricing |
|---|---|---|---|---|
sarvam-105b | 105B MoE. Complex reasoning, agentic tasks, long documents. | 131K | Yes | $0.1563 in / $0.625 out per 1M |
sarvam-105b-conversations | 105B MoE tuned for conversation and voice. Tool calling. | 32K | Yes | $0.1563 in / $0.625 out per 1M |
gpt-4o | Multimodal text + vision. | 128K | No | $2.604 in / $10.42 out per 1M |
gemini-3.8-flash | Fast multimodal flagship. Thinking low/medium/high. | 1M | No | $1.563 in / $7.813 out per 1M |
gemini-3.7-flash | Fast multimodal. Thinking low/medium/high. | 1M | No | $1.563 in / $7.813 out per 1M |
gemini-3.6-flash | Fast multimodal. Thinking minimal→high. | 1M | No | $1.563 in / $7.813 out per 1M |
gemini-3.5-flash | Balanced multimodal workhorse. | 1M | No | $1.563 in / $9.375 out per 1M |
gemini-3.5-flash-lite | Cheapest 1M-context Gemini. | 1M | No | $0.3125 in / $2.604 out per 1M |
gemini-3.1-pro-preview | Reasoning-heavy Gemini tier. | 1M | No | $2.083 in / $12.5 out per 1M |
gemini-3.1-flash-lite | Low-cost multimodal, tool use. | 1M | No | $0.2604 in / $1.563 out per 1M |
gpt-4.1 | Long-context multimodal. Strong instruction following. | 300K | No | $2.083 in / $8.333 out per 1M |
gpt-5-mini | Fast, affordable reasoning. | 400K | No | $0.2604 in / $2.083 out per 1M |
gpt-5.5 | Reasoning flagship. Vision, tools, prompt caching. | 1.05M | No | $5.208 in / $31.25 out per 1M |
gpt-5.6-sol | Frontier model for complex professional work. Vision, reasoning, tools. | 1.05M | No | $4.167 in / $20.83 out per 1M |
gpt-5.6-terra | Balances intelligence and cost. Vision, reasoning, tools. | 1.05M | No | $2.083 in / $12.5 out per 1M |
gpt-5.6-luna | Cost-sensitive, high-volume workloads. Vision, reasoning, tools. | 1.05M | No | $0.2083 in / $1.25 out per 1M |
gpt-6-sol | Frontier reasoning for enterprise agents, coding and complex knowledge work. Vision, reasoning, tools. | 1.05M | No | $2.083 in / $10.42 out per 1M |
gpt-6-luna | Efficient GPT-6 for high-volume, cost-sensitive workloads. Vision, reasoning, tools. | 1.05M | No | $0.1042 in / $0.5208 out per 1M |
gpt-6.1-sol | Near-Astra performance for complex coding, computer use and professional work at a lower cost. Vision, reasoning, tools. | 1.05M | No | $2.083 in / $10.42 out per 1M |
claude-opus-5-5 | Anthropic Claude Opus 5.5 for long-running agentic coding and knowledge work. 128K max output. Vision, reasoning, tools. Pro plan and up. | 1M | No | $4.167 in / $20.83 out per 1M ($0.2083 cached input) |
claude-sonnet-5-5 | Anthropic Claude Sonnet 5.5, the best combination of speed and intelligence. 128K max output. Vision, reasoning, tools. Pro plan and up. | 1M | No | $2.083 in / $10.42 out per 1M ($0.1042 cached input) |
claude-haiku-5-5 | Anthropic Claude Haiku 5.5 for high-volume, latency-sensitive tasks such as classification, extraction and routing. 128K max output. Vision, reasoning, tools. Pro plan and up. | 1M | No | $0.1042 in / $0.5208 out per 1M ($0.01042 cached input); prompts over 100K tokens (cached tokens included) bill $0.5208 / $2.604 ($0.05208 cached) |
grok-4.3 | xAI Grok 4.3. Reasoning + vision (images of at least 512 pixels in total; smaller returns 400). Reasoning tokens bill as output. | 200K | No | $3.646 in / $15.63 out per 1M |
DeepSeek-V4-Pro | Flagship DeepSeek reasoning. Tools. Text-only. | 1M | No | $1.813 in / $4.125 out per 1M |
DeepSeek-V4-Flash | Fast, affordable DeepSeek reasoning. Tools. Text-only. | 1M | No | $0.4583 in / $1.375 out per 1M |
kimi-k2.5 | Strong on coding and math. Vision. | 262K | Yes | $0.8438 in / $4.219 out per 1M |
kimi-k2.6 | Improved reasoning and coding over K2.5. Vision. | 262K | Yes | $1.333 in / $5.625 out per 1M |
kimi-k2.7-code | 1T-param agentic coding. Vision + tools. | 262K | Yes | $1.333 in / $5.625 out per 1M |
kimi-k3.1-pro | Moonshot AI's most capable open-weight multimodal agentic model. Image input, tools, JSON mode, reasoning (low / high / max, always on). 128K default max output. Paid plans only. | 1M | No | $2.083 in / $10.42 out per 1M ($0.2083 cached input) |
glm-5.3 | Z.ai GLM 5.3. Tools + reasoning (low / high / max). Text-only. | 1M | No | $2.125 in / $6.667 out per 1M |
deepseek-v4.1-flash | DeepSeek V4.1 Flash. Tools + reasoning (low / high / max). Text-only. | 1M | No | $0.427 in / $1.771 out per 1M |
glm-5.3-flash | Z.ai GLM 5.3 Flash. Fast, low-cost. Vision + tools + reasoning (low / high / max). | 1M | No | $0.229 in / $0.760 out per 1M |
deepseek-v4.1-flash-1017 | DeepSeek V4.1 Flash, October 17 snapshot. Vision + tools + reasoning (off by none; default high). | 1M | No | $0.427 in / $1.771 out per 1M |
gpt-oss-120b | Open-weight 120B MoE. Reasoning-grade at lower cost. | 131K | Yes | $1.042 in / $4.167 out per 1M |
gemma-4-31b | Gemma 4 31B instruct. Fast with thinking off, vision, tools. | 128K | No | $0.3646 in / $0.9375 out per 1M |
deepgram-think-fast | Deepgram Think-Fast. Early access, built for voice agents. Tools + reasoning (none / minimal / low / medium / high). Text-only. | 32K | No | $0.3646 in / $0.9375 out per 1M |
deepgram-phonellm-mini | phoneLLM mini (Pipecat). Early access, open-weights, built for voice pipelines. Tools. Text-only. | 32K | No | $0.3646 in / $0.9375 out per 1M |
gpt-realtime | OpenAI flagship realtime speech-to-speech model — STT + reasoning + function calling + TTS in one. 10 concurrent. | 32K | No | $4.167 in / $16.67 out per 1M • $0.3906/min |
gpt-realtime-mini | Lowest-cost realtime — same single-model shape as gpt-realtime, ~3× cheaper. 20 concurrent. | 32K | No | $0.625 in / $2.5 out per 1M • $0.1229/min |
gpt-realtime-2 | Newest realtime with stronger tool calling. 32K input, 4K output. | 32K | No | $4.167 in / $25 out per 1M • $0.3906/min |
gpt-realtime-1.5 | Pinned 1.5 snapshot of gpt-realtime. Use when you want version stability. | 32K | No | $4.167 in / $16.67 out per 1M • $0.3906/min |
gpt-realtime-2.1 | Latest realtime speech-to-speech — better alphanumeric recognition, silence/noise + interruption handling. Voice-agent only. | 32K | No | $4.167 in / $25 out per 1M • $0.3906/min |
gpt-realtime-2.1-mini | Distilled, lower-cost realtime for faster voice interactions. Voice-agent only. | 32K | No | $0.625 in / $2.5 out per 1M • $0.1219/min |
gpt-live-1 | GPT Live speech-to-speech — audio + text in and out, function calling, 14 voices (default marin). No image or video input. Voice-agent only. | — | No | $0.05208/min (billed per second) |
Speech to Text (15 models)
| Model ID | Description | Context | Free | Pricing |
|---|---|---|---|---|
saaras:v4 | 23 languages (22 Indic + English). Five output modes: transcribe, translate, verbatim, transliterate, code-mix. | — | Yes | $0.3125 / hr |
gnani-prisma-v3.0 | India-first telephony STT. English + 9 Indian languages, automatic language detection, mixed-language speech, keyterm boosting. Real-time and batch. | — | No | $0.2813 / hr |
ink-whisper | Cartesia Ink Whisper — 100 languages including Hindi, Urdu and Tamil. Dynamic chunking reduces hallucination across pauses and silence. File transcription + streaming. | — | No | $0.1875 / hr |
ink-2 | Cartesia Ink 2 — top-ranked for voice agents (8% WER on AppTek's 14-accent call-centre benchmark, vs 10% Deepgram Flux and 12% ElevenLabs). Self-detects turns. Languages: en, fr, hi, ja, es. Voice sessions only — not available for file transcription. | — | No | $0.5625 / hr |
assemblyai-universal-3.6-pro | AssemblyAI Universal-3.6 Pro — newest streaming STT with built-in turn detection. 32 languages including Hindi, Marathi, Urdu and Indian English. Voice sessions only — not available for file transcription. | — | No | $0.4688 / hr |
inworld-stt-1 | Inworld STT 1 — low-cost streaming STT, English only. Voice sessions only — not available for file transcription. | — | No | $0.1563 / hr |
mai-transcribe-2 | Preview. Microsoft AI MAI-Transcribe-2 — file transcription in 60 languages with automatic detection and Hinglish code-switching, up to 2 hours per file. POST /v1/audio/transcriptions only: not available in voice sessions or translation. Not processed in India. | — | No | $0.375 / hr |
scribe_v2 | ElevenLabs Scribe v2 — state-of-the-art file transcription in 100 languages including Hindi, Bengali, Tamil, Telugu, Marathi, Gujarati, Kannada, Malayalam, Punjabi, Odia, Assamese and Urdu, automatic language detection. File transcription only. | — | No | $0.2292 / hr |
scribe_v2_medical | ElevenLabs Scribe v2 Medical — Scribe v2 fine-tuned for clinical audio, 35% fewer errors on it, same 100 languages and price. File transcription only. | — | No | $0.2292 / hr |
scribe_v2_realtime | ElevenLabs Scribe v2 Realtime — real-time recognition in 100 languages with automatic detection. Voice sessions only — not available for file transcription. | — | No | $0.4063 / hr |
microsoft-speech-stt | Microsoft Speech — real-time streaming STT, processed in India. 81 languages (148 locales) including Hindi, Bengali, Tamil, Telugu, Marathi, Gujarati, Kannada, Malayalam, Punjabi, Odia, Assamese, Urdu, Bhojpuri and Indian English; one language per call. Voice sessions only — not available for file transcription. | — | No | $1.0417 / hr |
whisper | 99 languages. Transcription + translation to English. | — | No | $0.4167 / hr |
gpt-4o-transcribe | Higher accuracy than Whisper. Streaming. | — | No | $0.4167 / hr |
gpt-4o-mini-transcribe | Cheaper, faster streaming transcription. | — | No | $0.25 / hr |
gpt-4o-transcribe-diarize | File transcription with speaker labels. Not available for voice sessions. | — | No | $0.4167 / hr |
Text to Speech (23 models)
| Model ID | Description | Voices | Free | Pricing |
|---|---|---|---|---|
bulbul:v3 | Indic TTS across 11 Indian languages. | 37 | Yes | $0.3125 / 10K chars |
bulbul:v4-flash | Low-latency Indic TTS: 9 Indian languages + Hinglish. | 222 | Yes | $0.2083 / 10K chars |
gnani-timbre-v3.0 | Gnani Timbre v3 — India-first neural TTS, 10 Indian languages + Hinglish, streaming, expressive audio tags. | 12 | No | $0.2813 / 10K chars |
sonic-3.6 | Cartesia Sonic 3.6 — most natural conversational TTS. 44 languages, native-quality Hindi + Hinglish, sub-90ms first audio. | Searchable library | No | $0.5208 / 10K chars |
deepgram-flux-tts | Deepgram Flux TTS — conversational, built for voice agents. 42 languages including Hindi, Tamil, Telugu, Kannada, Marathi, Gujarati, Bengali and Urdu; 32 are early access. | 266 | No | $0.4688 / 10K chars |
rime-coda | Rime Coda — 8 languages including Hindi, two native Hindi voices. Voice sessions only. | 13 | No | $0.5208 / 10K chars |
inworld-tts-2 | Inworld TTS-2 — 94 languages including 11 Indian languages. Voice sessions only. | 14 | No | $0.2604 / 10K chars |
inworld-tts-2-flash | Inworld TTS-2 Flash — faster, lower-cost TTS-2, same voices and languages. Voice sessions only. | 14 | No | $0.1563 / 10K chars |
fish-audio-drama-3 | Fish Audio Drama 3 — expressive, 13 languages, no Indian languages. Voice sessions only. | 7 | No | $0.5208 / 10K chars |
gradium-tts-beta | Gradium TTS beta — English, French, German, Spanish and Portuguese. Voice sessions only. | 13 | No | $0.50 / 10K chars |
microsoft-neural-tts | Microsoft Neural voices — neural TTS, processed in India. 13 languages including 12 Indian languages, plain text only. Voice sessions only. | 57 | No | $0.1563 / 10K chars |
gpt-4o-mini-tts | Steerable — takes an instructions field to direct tone. | 6 | No | $0.2083 / 10K chars |
mai-voice-2.1 | Preview. Microsoft AI MAI-Voice-2.1 — the most expressive MAI voice model, for narration and long-form audio. 23 languages, plain text only. POST /v1/audio/speech and voice agents. Not processed in India. | 97 | No | $0.2292 / 10K chars |
mai-voice-2.1-flash | Preview. Microsoft AI MAI-Voice-2.1-Flash — low latency, built for real-time voice agents. Same 97 voices and 23 languages, plain text only. POST /v1/audio/speech and voice agents. Not processed in India. | 97 | No | $0.1563 / 10K chars |
eleven_v4 | ElevenLabs Eleven v4 — most emotive, highest quality; 91 languages incl. 12 Indian languages; audio tags performed. 21 premade voices plus any ElevenLabs voice ID. | 21+ | No | $0.8333 / 10K chars |
eleven_v4_turbo | ElevenLabs Eleven v4 Turbo — v4 expressiveness tuned for live calls (~100 ms to first audio upstream); 91 languages. | 21+ | No | $0.4167 / 10K chars |
eleven_v3 | ElevenLabs Eleven v3 — emotionally rich, dramatic delivery; 85 languages incl. 11 Indian languages; audio tags; multi-speaker dialogue. | 21+ | No | $0.8333 / 10K chars |
eleven_v3_conversational | ElevenLabs Eleven v3 Conversational — v3 tuned for natural dialogue in agents (~165 ms upstream); 85 languages. | 21+ | No | $0.4167 / 10K chars |
eleven_multilingual_v2 | ElevenLabs Multilingual v2 — lifelike, most stable on long-form speech; 29 languages incl. Hindi and Tamil. | 21+ | No | $0.8333 / 10K chars |
eleven_flash_v2_5 | ElevenLabs Flash v2.5 — fast and affordable (~100 ms upstream); 32 languages incl. Hindi and Tamil. The recommended ElevenLabs voice for calls. | 21+ | No | $0.4167 / 10K chars |
eleven_turbo_v2_5 | ElevenLabs Turbo v2.5 — functionally equivalent to Flash v2.5 (which is faster); 32 languages. | 21+ | No | $0.4167 / 10K chars |
eleven_flash_v2 | ElevenLabs Flash v2 — ultra-low-latency English (~135 ms upstream). | 21+ | No | $0.4167 / 10K chars |
eleven_turbo_v2 | ElevenLabs Turbo v2 — functionally equivalent to Flash v2 (which is faster); English only. | 21+ | No | $0.4167 / 10K chars |
Image Generation (9 models)
| Model ID | Description | Free | Pricing |
|---|---|---|---|
flux-2-pro | Flux 2 Pro. Flagship BFL fidelity. | No | $0.1042 / image |
flux-1.1-pro | Flux 1.1 Pro. Fast, production-grade. | No | $0.05208 / image |
gpt-image-2.5-sunburst | Most capable generation + editing. Inpainting, quality tiers to max. | No | $0.2604 / image |
gemini-3.1-flash-lite-image | Low-cost text-to-image with reference edits. 1K resolution. | No | $0.035 / image |
gpt-image-2.5-flare | Fast, high-quality everyday generation. On-image text + edits. | No | $0.2604 / image |
gpt-image-2 | Accurate on-image text rendering. | No | $0.2604 / image |
gpt-image-1.5 | Precise image editing. Strong logo/face preservation. | No | $0.2604 / image |
nano-banana-2 | Fast multimodal image generation. | No | $0.06979 / image |
nano-banana-pro | Flagship typography and fidelity. | No | $0.1396 / image |
Embeddings (2 models)
| Model ID | Description | Dimensions | Free | Pricing |
|---|---|---|---|---|
text-embedding-3-small | Fast, low-cost embeddings. Best price/performance for large corpora. | 1536 | Yes | $0.02083 / 1M input tokens |
text-embedding-3-large | Highest-accuracy embeddings. | 3072 | Yes | $0.1354 / 1M input tokens |
Both accept 8,192-token inputs and support shortening the vector with dimensions. See Embeddings.
Tip
Filter programmatically — GET /api/v1/models?category=llm, ?category=stt, ?category=tts, ?category=image, ?category=embedding, or ?free=true for free-plan models only.
Model Selection
Pass the model ID in your request:
# Indic LLM
response = client.chat.completions.create(
model="sarvam-105b",
messages=[{"role": "user", "content": "Hello in Hindi"}]
)
# Indic LLM with thinking mode
response = client.chat.completions.create(
model="sarvam-105b",
messages=[{"role": "user", "content": "Solve this step by step"}],
extra_body={"reasoning_effort": "high"}
)
# First-party flagship model
response = client.chat.completions.create(
model="gpt-6-luna",
messages=[{"role": "user", "content": "Hello"}]
)The API automatically routes to the correct backend based on the model ID:
- Bare names (
kimi-k2.5,gpt-6-sol,DeepSeek-V4-Pro,gpt-oss-120b, …) → direct-routed or first-party sarvam-*prefix → Indic LLMssaaras:*/bulbul:*/deepgram-*/ image IDs → the matching speech or image backend
Every ID is a plain CallMissed ID with no vendor prefix — including models we deploy on demand.