Models

Low cost, strong quality, and fast generation, with excellent reference-to-video performance. Ideal for scenarios that require a balanced combination of output quality, capability, and speed.

Extremely fast generation at a low price, well suited for both image-to-video and text-to-video. It also performs quite well on simple r2va tasks, making it highly suitable for large-scale content generation scenarios.

Kimi K3.1 Pro is MoonshotAI’s flagship reasoning model with a 1M-token context window, built for long-horizon agent workflows and multimodal understanding with image and video input. It delivers strong performance in coding, 3D creation, and financial research, while supporting streaming output, structured generation, and tool calling with configurable reasoning effort.

Kimi K3.1 Pro Highspeed is the faster variant of Kimi K3.1 Pro, delivering comparable intelligence with approximately 2–3× higher output speed and up to a 1M-token context window. It is optimized for long-horizon reasoning and agentic workflows, with strong performance in coding, 3D creation, financial analysis, tool use, and multimodal understanding of text, images, and video.

MiniMax-M3.1-Flash is a frontier multimodal language model with a 1M-token context window, designed for agentic reasoning, tool use, coding, and long-context tasks. It supports text, image, and video inputs, and offers tunable thinking depth through effort levels to balance reasoning quality, latency, and token usage.

DeepSeek V4.1 Flash is a 552B-parameter MoE model built on a new causal encoder-decoder architecture, designed for higher capability, faster reasoning, higher throughput, and lower serving cost. It natively supports multimodal visual understanding and delivers flagship-level intelligence with significantly reduced KV cache requirements, making it well suited for high-throughput and cost-sensitive agentic workloads.

Atria Dawn is a next-generation agentic model designed for research and engineering workflows that require continuous environmental understanding, tool use, and multi-step task execution. It helps users turn open-ended problems into executable and verifiable results across tasks such as deep research, software creation, structured content delivery, and authorized cybersecurity analysis.

H3-Fast is an ultra-fast video generation model built on post-training optimization of the MiniMax H3 standard version. It outputs 480P and 768P videos up to 15 seconds long, and supports text-to-video, image-to-video, and multimodal reference-based video generation. It is optimized for faster, lower-cost batch video creation across e-commerce, advertising, and marketing workflows.

Jev 1.14 is the latest version of Jev, TypeSafe's structured decision model and the first of its System One models. System One models make fast, structured decisions for software, returning a typed answer rather than free-form text. It is suited for routing, classification, and other decision points inside an application where a fast, predictable answer matters more than generated prose.

Learn more in TypeSafe's docs: https://docs.typesafe.ai/models

Gemini 4 Argon is a frontier model from Google DeepMind, released on September 30, 2026. It is the first model in the Gemini 4 series and uses a new naming scheme. It is strong in coding, financial research, legal drafting, and autonomously patching cybersecurity vulnerabilities. It accepts text, image, video, and file inputs and has a 1-million-token context window. On benchmarks, it scores 77.9% on DeepSWE v1.1, ahead of Claude Opus 5.5 and GPT-6 Astra, and ranks first on the Vals Index with 68.90%.

LLaDA-Image is a competitive 6B-parameter open-source unified image generation and editing model. It is the 50-step base variant for high-quality text-to-image generation and instruction-guided editing, supporting reference-image editing and Chinese-English text rendering with strong visual detail, natural lighting, and coherent composition.

Muse Wonder is Meta's reasoning model for complex agentic tasks, with a 1M-token context window and support for text, image, video, audio, and PDF inputs.

Seedream 5.0 Flash is an image generation and editing model from ByteDance Seed. It is the fast, cost-efficient tier of the Seedream 5.0 family, suited for high-volume production and interactive editing workflows that need precise edits at low latency.

Mistral Large 4 is Mistral AI's newest flagship model. It is currently offered as an API public preview through Mistral Studio. The model uses a granular mixture-of-experts design and accepts image input natively, with a vision encoder of about 1.6 billion parameters and a context window officially listed at one million tokens. Its training data spans more than 160 languages, including every official EU language. Mistral plans to publish the open weights at the end of October, so organizations will be able to self-host, although the license terms have not been released. Independent benchmarks place its intelligence score comparable to some rival models, but its cost per task is relatively high, so users should test it on their own workloads before adopting it.

Claude Haiku 5.5 is Anthropic’s cheapest, fastest, and most capable small model to date, built for high-volume, cost-sensitive workloads. It handles summaries, compaction, database queries, classification, and subagent coding tasks with low latency, making it especially effective for live customer support, browser use, and other speed-sensitive applications.

Gemini Nano Banana 2.1 is an upgraded multimodal reasoning image model for production-grade image generation and editing, with stronger prompt adherence, multi-reference editing, and improved visual fidelity. It understands text, images, video, and PDF inputs, supports interleaved text and image generation, and can reason over complex spatial and compositional instructions before generating images. The model also supports grounding with Google Search and Google Image Search for better long-tail entity recognition, visual understanding, and output accuracy.

Ling 3.1 Flash is a hybrid reasoning mixture-of-experts model from inclusionAI, with 25B active parameters out of 560B total.

FLUX.3 Image is Black Forest Labs' flagship image generation and editing model. It handles text-to-image and multi-reference editing with up to 10 input images, and renders at fixed resolution tiers from 768 up to 4K with a selectable aspect ratio. Pricing is a flat per-image rate that scales with the chosen resolution tier.

GPT-6.1 Sol is a production-grade OpenAI model designed for agentic workloads, optimized for long-horizon and multi-step tasks such as continuous tool use, complex coding, computer use, and professional workflows. It delivers capabilities approaching GPT-6 Astra at significantly lower cost, including especially low cached-input pricing that makes it well suited for reusable-context agent systems and enterprise deployment scenarios.

Qwen-Audio-3.1-ASR-Flash is a large-scale model supporting efficient short speech recognition, designed for high-quality, multilingual, and cultural content transcription scenarios. The model supports multilingual and multi-regional Chinese dialect recognition, and is optimized for the rhythm, cadence, and classical Chinese expression of ancient Chinese poetry. It features context enhancement, punctuation prediction, and text normalization capabilities. This version also supports controllable ASR/AST output for multiple dialects. It includes native transcription and polishing capabilities, as well as industrial-grade multi-person dialogue role-based transcription capabilities.