- Text Generation
Mercury 2 is a proprietary, diffusion-based large language model (dLLM) from Inception designed for extremely fast reasoning and text generation with a long 128K-token context window.
Powered by OpenAI
GPT-4o Transcribe
- Text Generation
GPT-4o Transcribe is an OpenAI model specialized for converting audio into accurate, time-aligned text transcripts. It is notable for handling natural speech, varied accents, and real-world audio conditions with high reliability.
Providers Comparison
Speed
Output tokens per second · Higher is better
L
LLM API
S
SiliconFlow (FP8)
G
GMI
C
CoreWeave
N
Novita
D
DeepSeek
P
Parasail (FP8)
D
DeepInfra (FP4)
End-to-End Response Time
Seconds · Lower is better
L
LLM API
S
SiliconFlow (FP8)
G
GMI
C
CoreWeave
N
Novita
D
DeepSeek
P
Parasail (FP8)
D
DeepInfra (FP4)
Price
USD per 1M tokens (blended) · Lower is better
L
LLM API
S
SiliconFlow (FP8)
G
GMI
C
CoreWeave
N
Novita
D
DeepSeek
P
Parasail (FP8)
D
DeepInfra (FP4)
About the model
What is GPT-4o Transcribe?
GPT-4o Transcribe is a transcription-focused variant of OpenAI’s GPT-4o model designed to turn spoken audio into structured text. It is mainly used for tasks like meeting notes, call and podcast transcription, caption generation, and transforming voice recordings into searchable documents. It also supports workflows that combine transcription with light understanding, such as summarizing or tagging segments of speech. It belongs to the GPT-4o family of multimodal OpenAI models adapted for speech-to-text transcription workloads.
Model capabilities
5 Core Capabilities
-
Speech Transcription
Converts spoken audio into accurate, punctuated text transcripts, handling diverse speakers, accents, and recording conditions in real time.
-
Multilingual Transcription
Transcribes speech from multiple languages into text, preserving language-specific characters, names, and terminology where supported.
-
Conversation Transcripts
Generates structured transcripts for dialogues, meetings, and interviews, distinguishing speakers when metadata or channel separation is available.
-
Media Captioning
Produces text captions from audio tracks in videos or podcasts, supporting workflows for accessibility, search, and content indexing.
-
Streaming Monitoring
Supports near real-time transcription for live audio streams, enabling monitoring, compliance checks, and rapid downstream processing.
Use cases
6 Most Valuable Use Cases
- Real-time Speech Transcription
- Meeting and Call Notes
- Customer Support Call Logging
- Media Caption Generation
- Voice-based Workflow Automation
- Audio Data Preprocessing
Transparent pricing
Cost Comparison
LLM API offers the lowest per‑minute transcription cost with best‑in‑class latency and uptime.
| Provider | Region | Latency | Throughput | Uptime | Input ($/1M) | Output ($/1M) | Context |
|---|---|---|---|---|---|---|---|
| LLM API BEST | Global | ~150ms | ~120 min/s | 99.99% | ~$0.004/min | ~$0.004/min | ~4 hour audio |
| OpenAI | Global | ~400ms | ~40 min/s | 99.9% | $0.006/min | $0.006/min | ~3 hour audio |
| Azure OpenAI | US East | ~450ms | ~35 min/s | 99.9% | ~$0.007/min | ~$0.007/min | ~3 hour audio |
| Google Cloud (Speech-to-Text via Gemini) | Global | ~500ms | ~30 min/s | 99.9% | ~$0.010/min | ~$0.010/min | ~2 hour audio |
| Amazon Web Services (Transcribe-like) | US East | ~550ms | ~25 min/s | 99.9% | ~$0.014/min | ~$0.014/min | ~2 hour audio |
Performance benchmarks
Technical Specifications
| Metric | GPT-4o Transcribe (OpenAI) | Whisper v3 (OpenAI) | Deepgram Nova-2 (Deepgram) |
|---|---|---|---|
| Avg Latency | ~180ms | ~250ms | ~220ms |
| Languages Supported | ~100+ | ~90+ | ~60+ |
| Price per Minute | ~$0.006 | ~$0.006 | ~$0.010 |
| Max Duration per Request | ~60 min | ~60 min | ~300 min |
| Accuracy (WER, English) | ~6–8% | ~7–9% | ~8–11% |
| Real-time Streaming Support | Yes | Yes | Yes |
| Throughput | ~50× RT | ~30× RT | ~60× RT |
| Uptime SLA | ~99.9% | ~99.9% | ~99.9% |
30-day usage via LLM API
- 3.8B
- Audio minutes transcribed
- 27M
- API requests
- 7.4M
- Unique projects using GPT-4o Transcribe
- 99.9%
- Avg uptime
Architecture & Integration
Why Build on LLM.API?
One unified API. Every major model. Built-in reliability, cost control, and observability.
-
Intelligent AI Routing
Automatically route each request to the best model across providers based on latency, cost, and quality—without changing your integration or retraining clients.
One endpoint, any model -
Cost-Aware Orchestration
Use pricing-aware routing, quotas, and policies to keep spend predictable while still hitting quality and latency targets across models and clouds.
Control spend by design -
Automatic Fallback Flows
Define failover rules once and let LLM.API retry on alternate providers or models when timeouts, rate limits, or provider outages occur.
Resilience built in -
Full-Stack Observability
Get traces, metrics, and structured logs for every request so you can debug prompts, compare providers, and tune performance in production.
See every token -
Task-Level Abstractions
Describe tasks like chat, extraction, or tooling once and let LLM.API pick and tune models behind the scenes, simplifying integration and future upgrades.
Code to tasks, not models -
High-Throughput Batch
Submit large jobs as batches with automatic chunking, retries, and aggregation to slash costs and saturate throughput without writing glue code.
Scale jobs, not code
Decision guide
When to Use — When NOT to Use
Use it if...
- You need accurate transcription of short English audio clips into text for downstream processing.
- Your use case involves batch-transcribing meeting recordings or interviews for search and summarization.
- You need to convert user-uploaded voice notes into text for productivity or messaging apps.
- Your use case involves generating captions or subtitles from prerecorded video or podcast audio.
- You need a reliable OpenAI-native transcription model that integrates cleanly with other GPT-4o workflows.
- Your use case involves prototyping speech-to-text features without managing separate ASR infrastructure.
Avoid if...
- You need advanced spoken-language understanding, reasoning, or dialog beyond simple transcription of audio.
- Your workload requires ultra-low-latency streaming ASR for live captioning or voice assistants.
- You need high-quality transcription for many low-resource languages not well-covered by OpenAI models.
- Your workload requires detailed diarization, speaker identification, or complex audio event classification.
- You need offline, on-device transcription where sending audio to cloud services is impossible.
- Your workload requires tightly controlled, fully open-source ASR components for strict compliance constraints.
FAQ
Frequently Asked Questions
-
What is GPT-4o Transcribe?
GPT-4o Transcribe is an OpenAI GPT-4o-based model on LLM.API specialized for accurate, low-latency speech-to-text transcription.
-
What is GPT-4o Transcribe best used for?
GPT-4o Transcribe is best for real-time or batch transcription of meetings, calls, podcasts, and other audio into structured text.
-
What modalities does GPT-4o Transcribe support?
GPT-4o Transcribe accepts audio input and returns text output; it is not intended for direct image or video understanding.
-
How is GPT-4o Transcribe priced on LLM.API?
GPT-4o Transcribe is billed on LLM.API per unit of audio processed, typically metered in minutes or seconds rather than tokens.
-
What is the context window of GPT-4o Transcribe?
GPT-4o Transcribe effectively handles long audio segments, but downstream text usage is constrained by the GPT-4o text context window.
-
How fast is GPT-4o Transcribe in terms of latency?
GPT-4o Transcribe is optimized for low latency and can stream partial transcriptions for near real-time use cases.
-
How do I call GPT-4o Transcribe through LLM.API?
You invoke GPT-4o Transcribe by specifying the model name in LLM.API audio endpoints and sending your audio file or stream payload.
-
How does GPT-4o Transcribe compare to general GPT-4o chat models?
GPT-4o Transcribe focuses on audio-to-text accuracy and efficiency, while general GPT-4o chat models focus on multi-turn natural language reasoning.
-
Does GPT-4o Transcribe support multiple languages?
GPT-4o Transcribe supports multilingual transcription, but accuracy can vary by language and audio quality.
-
What are the main limitations of GPT-4o Transcribe?
GPT-4o Transcribe may struggle with heavy background noise, overlapping speakers, very low-quality audio, or highly domain-specific jargon.
Compare
Competitive Models
- Text Generation
all-mpnet-base-v2 is a widely used English sentence-embedding model from Sentence Transformers that maps text to 768-dimensional vectors for semantic similarity tasks. It is built on Microsoft’s…
- Text Generation
Sora 2 Pro is an OpenAI model name that has been mentioned publicly, but as of now OpenAI has not released authoritative technical details or documentation…