Models

GPT 6.1 Sol delivers near-Astra performance at lower cost for complex coding, computer use, professional work, and agentic tool use.

GPT 6 Luna is the fastest GPT-6 variant for high-volume chat, extraction, routing, and lightweight tools.

GPT 6 Sol balances GPT-6 reasoning quality with lower latency and cost for production agent workloads.

DeepSeek V4.1 Flash combines a one-million-token context window with configurable thinking and vision input.

GPT 6 Astra is OpenAI's flagship GPT-6 model for complex reasoning, coding, research, and long-running agents.

GLM 5.3 Flash is the faster multimodal GLM variant for coding, screenshots, and tool-driven workflows.

GLM 5.3 is Zhipu AI's flagship long-context model for advanced coding and autonomous agents.

Claude Opus 5 is Anthropic's advanced model for complex agentic coding and enterprise knowledge work.

Kimi K3 is moonshot's general-purpose language and agent model with a 1,047,000-token context window, vision and tool calling, and API streaming support.

GPT 5.6 Luna is the low-latency GPT-5.6 option for high-throughput chat and structured tasks.

GPT 5.6 Sol is the highest-capability GPT-5.6 variant for demanding coding and multimodal reasoning.

GPT 5.6 Terra is the balanced GPT-5.6 option for general-purpose agents, tools, and image understanding.

Claude Sonnet 5 combines fast responses with frontier intelligence for coding and everyday tool-heavy agents.

GLM 5.2 focuses on coding, structured generation, diagrams, and optional reasoning over long contexts.

Claude Fable 5 is Anthropic's long-horizon reasoning model for demanding analysis and autonomous agent workflows.

Nemotron Ultra is NVIDIA's large mixture-of-experts model for difficult reasoning and multi-step agent tasks.

Gemini 3.5 Flash targets fast, high-throughput multimodal workloads with built-in thinking support.

DeepSeek V4 Flash is the legacy Flash identifier now served by DeepSeek V4.1 Flash for fast reasoning.

DeepSeek V4 Pro is the higher-capability DeepSeek V4 endpoint for long-context reasoning and agentic coding.

GPT 5.5 is openai's general-purpose language and agent model with a 400,000-token context window, vision and tool calling, and API streaming support.

Kimi 2.6 is moonshot's general-purpose language and agent model with a 262,144-token context window, vision and tool calling, and API streaming support.

GPT 5.4 is openai's general-purpose language and agent model with a 400,000-token context window, vision and tool calling, and API streaming support.

GPT 5.4 Thinking is openai's deliberate reasoning and complex analysis model with a 400,000-token context window, vision and tool calling, and API streaming support.

Gemini 3.1 Flash-Lite is optimized for economical high-volume translation, extraction, and simple agent tasks.

Fast Google image generation and editing.

Mercury 2 is Inception Labs' diffusion language model for fast reasoning, tool use, and multimodal chat.

Gemini 3.1 Pro Preview is Google's advanced multimodal model for agentic workflows, complex reasoning, and coding.

GPT 5.3 Codex is openai's coding and software-engineering model with a 400,000-token context window, vision and tool calling, and API streaming support.

Kimi K2.5 is moonshot's deliberate reasoning and complex analysis model with a 256,000-token context window, vision and tool calling, and API streaming support.

GLM 4.7 is an efficient earlier-generation Zhipu model for coding, structured output, and tool use.

GPT 5.2 Codex is openai's coding and software-engineering model with a 400,000-token context window, vision and tool calling, and API streaming support.

Gemini 3 Flash Preview is Google's legacy Gemini 3 speed-oriented model for multimodal generation.

OpenAI image generation and editing.

GPT 5.2 is openai's general-purpose language and agent model with a 400,000-token context window, vision and tool calling, and API streaming support.

GPT 5.2 Thinking is openai's deliberate reasoning and complex analysis model with a 400,000-token context window, vision and tool calling, and API streaming support.

Ministral 3B is Mistral's compact model for local or high-volume classification, routing, and short responses.

Mistral Large 3 is an open-weight multimodal mixture-of-experts model for general-purpose enterprise workloads.

Claude Opus 4.5 is Anthropic's high-capability model for demanding coding, research, writing, and agent workflows.

Grok 4.1 non-reasoning is a legacy low-latency alias now redirected by xAI to Grok 4.3 with reasoning disabled.

GPT 5.1 is openai's general-purpose language and agent model with a 400,000-token context window, vision and tool calling, and API streaming support.

GPT 5.1 Codex is openai's coding and software-engineering model with a 400,000-token context window, vision and tool calling, and API streaming support.

Kimi K2 Thinking is moonshot's deliberate reasoning and complex analysis model with a 256,000-token context window, vision and tool calling, and API streaming support.

Claude Haiku 4.5 is Anthropic's fastest model for responsive chat, classification, coding, and image analysis.

OpenAI video generation.

Claude Sonnet 4.5 balances intelligence and speed for coding, writing, vision, and tool-based agents.

GPT 5 Codex is openai's coding and software-engineering model with a 400,000-token context window, vision and tool calling, and API streaming support.

GPT 5 is openai's general-purpose language and agent model with a 128,000-token context window, vision and tool calling, and API streaming support.

GPT 5 Mini is openai's low-latency and high-volume model with a 400,000-token context window, vision and tool calling, and API streaming support.

Claude Opus 4.1 is a previous-generation Anthropic flagship for deep analysis, coding, and long-form writing.

GPT OSS 120B is openai's general-purpose language and agent model with a 128,000-token context window, tool calling, and API streaming support.

GPT OSS 20B is openai's general-purpose language and agent model with a 131,072-token context window, tool calling, and API streaming support.

Gemini 2.5 Flash-Lite is Google's smallest 2.5 model for cost-sensitive classification and data processing.

Gemini 2.5 Flash is Google's hybrid reasoning model for low-latency multimodal applications.

GPT o4 Mini is openai's low-latency and high-volume model with a 128,000-token context window and API streaming support.

Ashna-X1 is AshnaAI's balanced default for everyday chat, web research, tool use, and image-aware workflows.

GPT 4.1 is openai's general-purpose language and agent model with a 1,047,576-token context window, vision and tool calling, and API streaming support.

GPT 4.1 Mini is openai's low-latency and high-volume model with a 1,047,576-token context window, vision and tool calling, and API streaming support.

GPT 4.1 Nano is openai's low-latency and high-volume model with a 1,047,576-token context window, vision and tool calling, and API streaming support.

Gemini 2.5 Pro is a one-million-context multimodal reasoning model that excels at coding and complex analysis.

Streaming speech-to-text for voice applications.

GPT o3 Mini is openai's low-latency and high-volume model with a 128,000-token context window and API streaming support.

GPT o1 is openai's deliberate reasoning and complex analysis model with a 128,000-token context window and API streaming support.

GPT 4o Mini is openai's low-latency and high-volume model with a 128,000-token context window, vision and tool calling, and API streaming support.

GPT 4o is openai's general-purpose language and agent model with a 128,000-token context window, vision and tool calling, and API streaming support.

Mixtral 8x22B is an open mixture-of-experts model for multilingual text generation, coding, and analysis.

Multilingual text-to-speech model.

Grok 4.3 is xAI's fast one-million-context model with image input, configurable reasoning, and tool calling.

Highest-accuracy Voyage cross-encoder reranker for refining retrieval results.

Fast and cost-effective Voyage reranker for latency-sensitive retrieval.

Code embeddings optimized for coding-agent and repository retrieval.

Contextualized chunk embeddings that encode each chunk with the surrounding document context.

Balanced general-purpose and multilingual retrieval embeddings in the shared Voyage 4 vector space.

Highest-quality Voyage general-purpose and multilingual retrieval embeddings.

Low-latency, low-cost retrieval embeddings in the shared Voyage 4 vector space.

Interleaved text, image, and video embeddings for rich multimodal retrieval.

Previous-generation multilingual reranker with instruction-following support.

Previous-generation low-latency multilingual and instruction-aware reranker.

Previous-generation contextualized chunk embeddings for document-aware retrieval.

Previous-generation general-purpose and multilingual retrieval embeddings.

Previous-generation retrieval embeddings optimized for latency and cost.

Previous-generation high-quality general-purpose and multilingual retrieval embeddings.

Previous-generation code retrieval embeddings with flexible dimensions and quantization.

Previous-generation interleaved text and image embeddings for visual retrieval.

General-purpose and multilingual retrieval embeddings with a fixed 1024-dimensional output.

Cost-efficient general-purpose retrieval embeddings with a fixed 512-dimensional output.

Multilingual retrieval and RAG embeddings supporting major global languages.

Domain-specific embeddings optimized for financial retrieval and RAG.

Instruction-tuned embeddings for retrieval, classification, clustering, and similarity.

Domain-specific embeddings optimized for legal and long-context retrieval.

High-quality OpenAI text embeddings.

Low-cost OpenAI text embeddings.

Second-generation semantic code search and code retrieval embeddings.

Legacy general-purpose embeddings balancing retrieval quality, latency, and cost.

Legacy general-purpose retrieval model with 1536-dimensional embeddings.