Claude Opus 5.5

Anthropic

For long-running agentic coding and knowledge work.

Claude Sonnet 5.5

Anthropic

The best combination of speed and intelligence.

Claude Fable 5.1

Anthropic

Claude Fable 5.1 is the current model in Anthropic's Fable family. It extends Claude Fable 5 with stronger long-running agentic coding, multistep research, and document, spreadsheet, and slide work. Anthropic recommends it for demanding reasoning and long-horizon agentic work, or when evals on Claude Opus 5 at higher effort still fall short.

GPT-6 Astra

OpenAI

GPT-6 Astra is OpenAI's most capable model, built for the hardest end-to-end work. OpenAI recommends it for complex reasoning, coding, computer use, research, and document creation. Reasoning effort supports low, medium, high, xhigh, and max.

Gemini 4 Argon

Google

Gemini 4 Argon is Google's most capable model, built for the hardest problems: long-horizon coding, agentic workflows, scientific, legal and financial analysis, and defensive security work. It reads text, images, audio, video and PDFs across a 1M-token context, thinks at low, medium or high effort, and calls tools reliably over many steps.

GPT-6.1 Sol

OpenAI

The GPT-6.1 update of Sol, the middle of the GPT-6 line: a million-token window at a fifth of Astra's price. Reasoning, tools and pictures, for work that does not need the dearest model in the family.

Claude Opus 5

Anthropic

Claude Opus 5 is the current model in Anthropic's Opus family, for complex agentic coding and enterprise work. Anthropic recommends starting with it for most workloads. It is a step-change improvement over Claude Opus 4.8, with the largest gains in deep reasoning, agentic and long-horizon tasks, and test-time compute scaling.

Claude Fable 5

Anthropic

Claude Fable 5 is a legacy model in Anthropic's Fable family, built for demanding reasoning and long-horizon agentic work. Claude Fable 5.1 is the current Fable model, and Anthropic recommends migrating to it for improved performance. Claude Fable 5 includes safety classifiers that can decline requests.

GPT-6 Sol

OpenAI

The middle of the GPT-6 line: the same million-token window as Astra, at a fifth of the price. Reasoning, tools and pictures, for work that does not need the dearest model in the family.

GPT-5.6 Sol

OpenAI

GPT-5.6 Sol is a flagship model in the GPT-5.6 family. It roughly corresponds to the unsuffixed model tier used in earlier GPT-5 families. The gpt-5.6 alias routes requests to GPT-5.6 Sol.

Claude Haiku 5.5

Anthropic

For high-volume, latency-sensitive tasks such as classification, extraction, and routing.

GPT-5.6 Terra

OpenAI

GPT-5.6 Terra is designed for workloads that balance intelligence and cost. It roughly corresponds to the mini model tier used in earlier GPT-5 families. Reasoning effort supports none, low, medium, which is the default, plus high, xhigh, and max.

Claude Opus 4.8

Anthropic

Claude Opus 4.8 is a legacy model in Anthropic's Opus family, released May 28, 2026. Claude Opus 5 is the current Opus model, and Anthropic recommends migrating to it for improved performance. Anthropic describes Claude Opus 5 as a step-change improvement over Claude Opus 4.8.

Gemini 3.8 Flash

Google

Gemini 3.8 Flash is Google's most intelligent Flash model, engineered for long-horizon software engineering, autonomous agents, and complex enterprise workflows, with the speed and cost efficiency of Flash. It is the current top Flash-tier option, and it accepts text, images, video, audio and PDF, returning text.

Claude Opus 4.7

Anthropic

Claude Opus 4.7 is a legacy model in Anthropic's Opus family, released April 16, 2026. It introduced the tokenizer used by the current Claude lineup. Claude Opus 5 is the current Opus model, and Anthropic recommends migrating to it for improved performance.

Gemini 3.7 Flash

Google

Gemini 3.7 Flash is an iteration in the Gemini 3 series of natively multimodal reasoning models. Google now lists it as a previous-generation Flash model, positioned for complex coding, agentic workflows, and reliable multi-step execution. It supports thinking at low, medium and high settings.

GPT-5.4

OpenAI

GPT-5.4 is a flagship model for complex professional work, and OpenAI presents it as a more affordable model for coding and professional work. Reasoning effort supports none, which is the default, plus low, medium, high, and xhigh.

GPT-5.5

OpenAI

GPT-5.5 is a flagship model for the most complex professional work, described by OpenAI as a new class of intelligence for coding and professional work. Reasoning effort supports none, low, medium, which is the default, plus high and xhigh.

Claude Sonnet 5

Anthropic

Claude Sonnet 5 is the next generation of Anthropic's Sonnet model family, offering the best combination of speed and intelligence in the current lineup. It is a drop-in upgrade for Claude Sonnet 4.6, with adaptive thinking on by default and manual extended thinking no longer accepted.

GPT-6 Luna

OpenAI

The cheapest of the GPT-6 line, built for volume: a million-token window at ten cents a million. For classifying, extracting and summarising at scale rather than for hard reasoning.

GPT-5.6 Luna

OpenAI

GPT-5.6 Luna is designed for cost-sensitive, high-volume workloads. It roughly corresponds to the nano model tier used in earlier GPT-5 families. Reasoning effort supports none, low, medium, which is the default, plus high, xhigh, and max.

Gemini 3.6 Flash

Google

Gemini 3.6 Flash provides sustained frontier-level intelligence optimized for real-world tasks at higher speed and lower cost. Designed for the agentic era, it excels at code generation, agentic execution, and spatial reasoning. Google now lists it as a previous-generation Flash model for general agentic and everyday tasks.

Gemini 3.5 Flash

Google

Gemini 3.5 Flash provides sustained frontier-level intelligence optimized for real-world tasks at higher speed and lower cost. Designed for the agentic era, it excels at sub-agent deployment, multi-step workflows, and long-horizon tasks at scale. Google now lists it as a legacy Flash model for routine, high-throughput workloads. It thinks before it answers, and that thinking counts toward max_tokens: allow at least 1024 tokens for a short answer and 4096 or more (or reasoning effort low) for a longer one.

GPT-5.3 Codex

OpenAI

GPT-5.3 Codex is optimized for agentic coding tasks in Codex or similar environments. OpenAI calls it the most capable agentic coding model to date. It supports low, medium, high, and xhigh reasoning effort settings, and it has a dedicated prompting guide.

Claude Opus 4.6

Anthropic

Claude Opus 4.6 is a legacy model in Anthropic's Opus family, released February 5, 2026. It supports adaptive thinking, and manual extended thinking is deprecated on it. Claude Opus 5 is the current Opus model, and Anthropic recommends migrating to it for improved performance.

GPT-5.2

OpenAI

GPT-5.2 is OpenAI's previous flagship model for complex professional work, with configurable reasoning effort. Reasoning effort supports none, which is the default, plus low, medium, high, and xhigh. OpenAI recommends using the latest model, GPT-6 Astra.

Claude Sonnet 4.6

Anthropic

Claude Sonnet 4.6 is a legacy model in Anthropic's Sonnet family, released February 17, 2026. It supports adaptive thinking, and manual extended thinking is deprecated on it. Claude Sonnet 5 is the current Sonnet model, and Anthropic recommends migrating to it for improved performance.

Claude Opus 4.5

Anthropic

Claude Opus 4.5 is a legacy model in Anthropic's Opus family, released November 24, 2025. It uses manual extended thinking rather than the adaptive thinking of later Opus models. Claude Opus 5 is the current Opus model, and Anthropic recommends migrating to it for improved performance.

GPT-5.1

OpenAI

GPT-5.1 is a flagship model for coding and agentic tasks with configurable reasoning and non-reasoning effort. Reasoning effort supports none, which is the default, plus low, medium, and high. OpenAI publishes separate GPT-5.1 model guidance for it.

GPT-5.4 mini

OpenAI

GPT-5.4 mini brings the strengths of GPT-5.4 to a faster, more efficient model designed for high-volume workloads. OpenAI calls it its strongest mini model yet for coding, computer use, and subagents. Reasoning effort supports none by default, plus low, medium, high, and xhigh.

Gemini 3.5 Flash-Lite

Google

Gemini 3.5 Flash-Lite is a low-latency, cost-effective multimodal model optimized for high-throughput execution on subagent tasks and document parsing. It is the fastest and most cost-effective model in the 3.5 generation. Google recommends it for high-volume agentic workflows, simple data extraction, and applications where latency and API cost are the primary constraints.

GPT-5.4 nano

OpenAI

GPT-5.4 nano is the cheapest GPT-5.4-class model, designed for tasks where speed and cost matter most, such as classification, data extraction, ranking, and sub-agents. Reasoning effort supports none, which is the default, plus low, medium, high, and xhigh.

Claude Sonnet 4.5

Anthropic

Claude Sonnet 4.5 is a legacy model in Anthropic's Sonnet family, released September 29, 2025. It uses manual extended thinking and does not support the effort parameter. Claude Sonnet 5 is the current Sonnet model, and Anthropic recommends migrating to it for improved performance.

Claude Haiku 4.5

Anthropic

Claude Haiku 4.5 is the fastest model in Anthropic's current lineup, with near-frontier intelligence. It uses manual extended thinking rather than the adaptive thinking of the larger current models, and does not support the effort parameter. The ID claude-haiku-4-5 is an alias that resolves to this pinned snapshot.

Gemini 2.5 Pro

Google

Gemini 2.5 Pro is Google's state-of-the-art thinking model, capable of reasoning over complex problems in code, math, and STEM. It also analyzes large datasets, codebases, and documents using long context. It is the most advanced model in the Gemini 2.5 family, above Flash and Flash-Lite.

Gemini 2.5 Flash

Google

Gemini 2.5 Flash is Google's best price-performance model in the Gemini 2.5 family, offering well-rounded capabilities. It is recommended for large scale processing, low-latency, high-volume tasks that require thinking, and agentic use cases. It accepts text, images, video and audio, and returns text.

o1-pro

OpenAI

o1-pro is the version of o1 that uses more compute to think harder and provide consistently better answers. The o1 series is trained with reinforcement learning to think before answering and perform complex reasoning. o1-pro is available in the Responses API only.

Gemini 2.5 Flash-Lite

Google

Gemini 2.5 Flash-Lite is the fastest and most cost-efficient multimodal model in the Gemini 2.5 family. Google recommends it for high-volume classification, simple data extraction, and extremely low-latency applications where budget and speed are the primary constraints. It accepts text, images, video, audio and PDF, and returns text.

GPT-4o mini

OpenAI

GPT-4o mini is a fast, affordable small model for focused tasks. It accepts both text and image inputs and produces text outputs, including Structured Outputs. It is ideal for fine-tuning, and outputs from a larger model like GPT-4o can be distilled into it for similar results at lower cost and latency.

GPT-4o

OpenAI

GPT-4o ('o' for 'omni') is OpenAI's versatile, high-intelligence flagship model. It accepts both text and image inputs and produces text outputs, including Structured Outputs. OpenAI describes it as the best model for most tasks and its most capable model outside the o-series.

GPT Image 1

OpenAI

GPT Image 1 is a natively multimodal language model that accepts both text and image inputs, and produces image outputs. OpenAI's documentation lists it as a previous generation image model, with GPT Image 2 as the current one.

Gemini Nano Banana 2.1

Google

Gemini Nano Banana 2.1 generates and edits images from text, image and video inputs, with thinking support. It supports 1K, 2K and 4K output and up to 14 reference images.

Gemini 3.5 Transcribe

Google

Gemini 3.5 Transcribe is a speech-to-text model based on Gemini's audio understanding capabilities. It provides low-latency transcription with utterance-based language detection, speaker diarization, word-level timestamps, smart transcription, and custom vocabulary speech biasing. Google offers it in a version for audio files and a live version for streaming.

Gemini 3.1 Flash-Lite Image

Google

Gemini 3.1 Flash-Lite Image, which Google calls Nano Banana 2 Lite, is the efficiency specialist of the image generation family, targeting sub-2 second end-to-end latency. It is built for high-volume interactive developer use cases and real-time consumer applications, and for fast multi-turn local edits such as color swaps and background changes.

GPT-5.5 Pro

OpenAI

GPT-5.5 Pro is the version of GPT-5.5 that uses more compute to think harder and provide consistently better answers. It is available for Responses API requests, including through the Batch API. Reasoning effort supports medium, high, which is the default, and xhigh.

Gemini Embedding 2

Google

Gemini Embedding 2 is the first multimodal embedding model in the Gemini API, mapping text, images, video, audio, and documents into one unified embedding space across over 100 languages. Google recommends it for cross-modal semantic search, document retrieval, and recommendation systems. Upgrading from the earlier embedding model requires re-embedding all existing data.

GPT-5.4 Pro

OpenAI

GPT-5.4 Pro is the version of GPT-5.4 that uses more compute to think harder and provide consistently better answers. It is available in the Responses API only, and reasoning effort supports medium, which is the default, plus high and xhigh. Some requests may take several minutes to finish.

Gemini 3.1 Flash-Lite

Google

Gemini 3.1 Flash-Lite is a low-latency, cost-effective multimodal model optimized for high-frequency, lightweight tasks. Google recommends it for high-volume translation, audio transcription, lightweight data extraction, document summarization, and as a classifier that routes requests to more capable models. It does not support image generation, audio generation, or computer use.

Veo 3.1 Lite

Google

Veo 3.1 Lite is a high-efficiency, developer-first video model providing video generation, editing, and cinematic control. Google positions it for developers and enterprises building high-volume video applications and workflows at scale. Unlike the other Veo 3.1 models, it does not support 4K output or the extension feature.

Gemini 3.1 Flash Image

Google

Gemini 3.1 Flash Image, which Google calls Nano Banana 2, provides high-quality image generation and conversational editing at low latency. It is the high-efficiency counterpart to Gemini 3 Pro Image, optimized for speed and high-volume developer use cases.

Gemini 3.1 Pro

Google

Gemini 3.1 Pro is built to refine the performance and reliability of the Gemini 3 Pro series, with better thinking, improved token efficiency, and a more grounded, factually consistent experience. It is optimized for software engineering behavior and for agentic workflows that require precise tool usage and reliable multi-step execution.

GPT-5.2 Pro

OpenAI

GPT-5.2 Pro is OpenAI's previous pro model for complex professional work. It is available in the Responses API only and supports medium, high, and xhigh reasoning effort. OpenAI recommends GPT-5.5 Pro as the latest pro model.

Gemini 2.5 Pro TTS

Google

Gemini 2.5 Pro TTS is Google's premium engine for studio-quality speech synthesis, sitting above the Flash text-to-speech model. It is recommended for long-form content, professional narrations, and complex creative workflows that require the highest level of vocal clarity and natural prosody.

Gemini 2.5 Flash TTS

Google

Gemini 2.5 Flash TTS generates speech from text through Vertex AI, with Kore among its prebuilt voices. Speech is available as WAV or PCM.

Gemini 3 Pro Image

Google

Gemini 3 Pro Image, which Google calls Nano Banana Pro, is a reasoning-driven engine for professional-grade image editing and generation with studio-quality precision and advanced creative control. Google recommends it for complex graphic design, high-fidelity product mockups, and factual data visualizations that need accurate text rendering and grounding in Google Search.

GPT-5 Pro

OpenAI

GPT-5 Pro is the version of GPT-5 that uses more compute to think harder and provide consistently better answers. It is available in the Responses API only, and it defaults to and only supports high reasoning effort. Some requests may take several minutes to finish.

GPT-5

OpenAI

GPT-5 is OpenAI's previous model for coding, reasoning, and agentic tasks across domains, with configurable reasoning effort. Reasoning effort supports minimal, low, medium, and high. OpenAI recommends using the latest model, GPT-6 Astra, instead.

GPT-5 mini

OpenAI

GPT-5 mini is a faster, more cost-efficient version of GPT-5. It is good for well-defined tasks and precise prompts. For most new low-latency, high-volume workloads OpenAI recommends starting with GPT-5.6 Terra.

GPT-5 nano

OpenAI

GPT-5 nano is the fastest, most cost-efficient version of GPT-5. It is good for summarization and classification tasks. For most new speed- and cost-sensitive workloads OpenAI recommends starting with GPT-5.6 Luna.

o3-pro

OpenAI

o3-pro is the version of o3 that uses more compute to think harder and provide consistently better answers. The o-series is trained with reinforcement learning to think before answering and perform complex reasoning. o3-pro is available in the Responses API only, and some requests may take several minutes.

o4-mini

OpenAI

o4-mini is OpenAI's latest small o-series model. It is optimized for fast, effective reasoning with exceptionally efficient performance in coding and visual tasks. It accepts text and image input and returns text. o4-mini is succeeded by GPT-5 mini.

o3

OpenAI

o3 is a well-rounded and powerful model across domains, and it sets a new standard for math, science, coding, and visual reasoning tasks. It also excels at technical writing and instruction-following. Use it to think through multi-step problems across text, code, and images. o3 is succeeded by GPT-5.

GPT-4.1

OpenAI

GPT-4.1 is OpenAI's smartest non-reasoning model. It excels at instruction following and tool calling, with broad knowledge across domains and low latency because there is no reasoning step. OpenAI recommends starting with GPT-5 for more complex tasks.

GPT-4.1 mini

OpenAI

GPT-4.1 mini is a smaller, faster version of GPT-4.1. It excels at instruction following and tool calling, with low latency and no reasoning step. OpenAI recommends starting with GPT-5 mini for more complex tasks.

GPT-4.1 nano

OpenAI

GPT-4.1 nano is the fastest and most cost-efficient version of GPT-4.1. It excels at instruction following and tool calling, with low latency and no reasoning step. OpenAI recommends starting with GPT-5 nano for more complex tasks.

o3-mini

OpenAI

o3-mini is a small model alternative to o3. It provides high intelligence at the same cost and latency targets as o1-mini. It supports key developer features such as Structured Outputs, function calling, and the Batch API.

o1

OpenAI

o1 is OpenAI's previous full o-series reasoning model. The o1 series of models are trained with reinforcement learning to perform complex reasoning. They think before they answer, producing a long internal chain of thought before responding to the user.

Veo 3.1 Fast

Google

Veo 3.1 Fast is the faster variant of the Veo 3.1 video generation model. Google's Veo documentation lists it with the same core capabilities as standard Veo 3.1, including text and image input, video output with native audio, 4K resolution, and video extension.

Veo 3.1

Google

Veo 3.1 is a cinematic engine designed for high-end creative storytelling and experimental video production. Google recommends it for professional-grade 4K output, natively synchronized audio generation, and complex camera movements that require the highest level of temporal consistency and artistic control. It generates video from text and image input.

Text Embedding 3 Large

OpenAI

Text Embedding 3 Large is OpenAI's most capable embedding model for both English and non-English tasks. Embeddings are numerical representations of text used to measure relatedness, and they are useful for search, clustering, recommendations, anomaly detection and classification. It serves the Embeddings endpoint and is the larger of the two version 3 models.

GPT Image 2.5 Flare

OpenAI

GPT Image 2.5 Flare is one of the two GPT Image 2.5 variants available here. A prompt and any reference pictures go in and an image comes back; it reads a 32k context and shares the family's limits and price with Sunburst beside it.

GPT Image 2.5 Sunburst

OpenAI

GPT Image 2.5 Sunburst is one of the two GPT Image 2.5 variants available here. A prompt and any reference pictures go in and an image comes back; it reads a 32k context and shares the family's limits and price with Flare beside it.

Gemini 2.5 Flash Live

Google

Gemini 2.5 Flash Live is Google's model for conversations held out loud: it listens to text, sound and video and answers in both speech and text. It carries a 131k context with tool calling, and like every realtime model it is reached over a WebSocket held open for the length of the conversation rather than on the chat endpoint.

GPT Realtime 2.1

OpenAI

GPT Realtime 2.1 is the newest of the live-voice models, keeping the 128k context and 32k answers of GPT Realtime 2. It listens to sound, text and pictures, answers out loud over a WebSocket, and calls tools mid-conversation without breaking the session.

GPT Realtime 2.1 Mini

OpenAI

GPT Realtime 2.1 Mini is the light tier of the newest live-voice generation — the 128k context and 32k answers of GPT Realtime 2.1 at roughly a sixth of its price. Sound, text and pictures in, speech out, over the realtime WebSocket endpoint.

GPT Image 2

OpenAI

GPT Image 2 is OpenAI's state-of-the-art image generation model for fast, high-quality image generation and editing. It supports flexible image sizes and high-fidelity image inputs, taking text and image input and returning images. It is the current image model, replacing GPT Image 1 and GPT Image 1.5.

GPT Realtime 2

OpenAI

GPT Realtime 2 widens the live-voice family to a 128k context and answers of up to 32k — four times what the first generation held. It takes sound, text and pictures, answers in speech over a WebSocket, and calls tools in the middle of a conversation.

Text Embedding Ada 002

OpenAI

Text Embedding Ada 002 is an improved, more performant version of OpenAI's original ada embedding model. Embeddings are numerical representations of text used to measure relatedness, and they are useful for search, clustering, recommendations, anomaly detection and classification. OpenAI lists it as an older embedding model, preceding the version 3 embedding models.

GPT Audio 1.5

OpenAI

GPT Audio 1.5 is the newer release of the GPT Audio family — a chat model that answers aloud and hands back the transcript of what it said. It keeps the family's 128k context, 16k answers, tool calling and price, and runs on the ordinary chat endpoint rather than over a WebSocket.

GPT Image 1.5

OpenAI

GPT Image 1.5 is described by OpenAI as its previous image generation model, with better instruction following and adherence to prompts. It accepts text and image input and can return both images and text. GPT Image 2 is the current model in this line.

GPT Realtime 1.5

OpenAI

GPT Realtime 1.5 is the second release of the live-voice family, keeping the 32k context, vision and tool calling of GPT Realtime. Like every model in this family it is reached over a WebSocket rather than on chat completions, and a call sent to the wrong door is refused with the right one named.

TTS-1

OpenAI

TTS-1 is a text-to-speech model that converts text into natural sounding spoken audio. It is optimized for realtime text-to-speech use cases, where TTS-1 HD is optimized for quality instead. It is used with the Speech endpoint in the Audio API.

TTS-1 HD

OpenAI

TTS-1 HD is a text-to-speech model that converts text into natural sounding spoken audio. It is optimized for high quality text-to-speech use cases, where the plain TTS-1 model is optimized for speed. It is used with the Speech endpoint in the Audio API.

Whisper

OpenAI

Whisper is a general-purpose speech recognition model, trained on a large dataset of diverse audio. It is a multitask model that handles multilingual speech recognition, translation and language identification. It serves the transcription and translation endpoints of the Audio API, and it is the one transcription model that takes audio alone, with no text prompt.

GPT Image 1 mini

OpenAI

GPT Image 1 mini is a cost-efficient version of GPT Image 1. It is a natively multimodal language model that accepts both text and image inputs, and produces image outputs. It belongs to the GPT Image 1 generation rather than the current GPT Image 2.

ChatGPT

OpenAI

ChatGPT points to the latest Instant model currently used in the ChatGPT product, and the underlying model snapshot is updated regularly. OpenAI recommends GPT-6 Astra for production API usage instead of this alias.

GPT-4o Transcribe

OpenAI

GPT-4o Transcribe is a speech-to-text model that uses GPT-4o to transcribe audio. It offers improvements to word error rate and better language recognition and accuracy compared to the original Whisper models. Unlike Whisper it accepts text input as well as audio, so a prompt can guide the transcript.

GPT-4o Transcribe Diarize

OpenAI

GPT-4o Transcribe Diarize is an automatic speech recognition model with built-in speaker diarization. It attributes segments of audio to distinct speakers in a conversation, so a transcript records who spoke when. It is available only through the Transcription API.

GPT Audio

OpenAI

GPT Audio is a chat model that answers in a voice: text or a recording goes in, and both the spoken answer and a transcript of it come back. Unlike the realtime family it runs on the ordinary chat endpoint, with a 128k context, answers up to 16k and tool calling.

GPT Realtime

OpenAI

GPT Realtime is OpenAI's model for live voice sessions — sound, text and pictures go in, speech and text come back, over a WebSocket held open while the conversation lasts. It carries a 32k context with vision and tool calling, and answers only on the realtime endpoint, never on chat completions.

GPT Transcribe

OpenAI

GPT Transcribe is a speech-to-text model for completed audio files, streamed file transcripts, and committed turns in Realtime sessions over WebSocket. It supports unstructured context, keyword hints and multiple language hints, which improve transcription of domain terms, multilingual audio and code-switching.

Omni Moderation

OpenAI

Omni Moderation is a moderation model that identifies potentially harmful content in text and images. It classifies the input it is given rather than generating text. It is OpenAI's most capable moderation model, and it accepts images as input as well as text.

GPT-4o mini Transcribe

OpenAI

GPT-4o mini Transcribe is a speech-to-text model that uses GPT-4o mini to transcribe audio. It offers improvements to word error rate and better language recognition and accuracy compared to the original Whisper models. It is the smaller sibling of GPT-4o Transcribe and takes audio and text input.

GPT-4o mini TTS

OpenAI

GPT-4o mini TTS turns written text into speech and does nothing else: text in, a recording out, no conversation. It is billed by token rather than by character — about 1900 audio tokens per thousand characters — so the cost of a long read can be worked out before it is spoken.

GPT Audio Mini

OpenAI

GPT Audio Mini is the light tier of the GPT Audio family: the same spoken answers, the same 128k context and the same chat endpoint, at roughly a quarter of the price. It is the one to reach for when the volume matters more than the hardest conversations do.

GPT Realtime Mini

OpenAI

GPT Realtime Mini is the light tier of the live-voice family: the same WebSocket sessions, vision and tool calling as GPT Realtime, with a 32k context, at roughly a sixth of the price. Built for conversations you hold many of rather than long ones.