Models

Voxtral TTS 3 is Mistral's text-to-speech model. It converts text into natural-sounding speech, offers a library of preset voices, supports zero-shot voice cloning, and accepts emotion and pace tags in the input to control delivery. It streams audio with low time-to-first-audio and suits voice agents, narration, and multilingual assistants.

Voxtral Small Transcribe 3 is Mistral's flagship speech-to-text model. It takes audio input and returns a transcript with timestamps and speaker diarization, and supports context biasing for domain-specific vocabulary. Suited for high-accuracy multilingual transcription of meetings, interviews, media, and support calls.

Voxtral Mini Transcribe 3 is Mistral's compact speech-to-text model. It takes audio input and returns a transcript with timestamps and speaker diarization. It also supports context biasing for domain terms. It is small enough to self-deploy in resource-constrained environments, and fits meeting notes, call analytics, and voice-agent pipelines where cost matters.

MiniMax-M3.1-Flash is a multimodal model from MiniMax. Compared with M3, it delivers significant improvements across multiple benchmarks, particularly in complex agentic tasks, coding, and cybersecurity. It is built for complex, multi-step co-work automation, understanding real codebases, fixing bugs, and building features, and visual understanding and creation, including frontend and presentation creation. Open weights are planned.

DeepSeek V4.1 Flash 1017 is the October 17, 2026 snapshot of DeepSeek V4.1 Flash, served by DeepSeek under the deepseek-flash API model name. V4.1 Flash is a sparse mixture-of-experts model from DeepSeek, and the first built on the company's Causal Encoder-Decoder (CED) architecture. It activates 8B parameters on input and 16B on output from a 552B-parameter backbone, an asymmetric split that keeps per-token compute low relative to the model's total size. Image understanding is native to the architecture, with visual and text embeddings trained jointly from the start of pre-training rather than added afterward as in the earlier experimental V4 Flash Vision Exp. It is suited for coding, terminal, and computer-use agents, along with long-horizon tasks that must run to completion across many steps and long-context analysis. Compressed KV caching cuts cache memory to roughly a quarter of the previous Flash generation, significantly reducing costs on agentic workloads. DeepSeek positions it as the cost-efficient tier of the V4.1 family and reports that it exceeds V4 Pro on performance, speed, and task completion time.

Glimpse is a multimodal visual reasoning model from Elorian, built for visual understanding and complex visual reasoning in the physical world. It reasons directly over images and video rather than converting them to text first, so it can find small details in dense scenes, track events across video frames, and ground its answers in what is actually on screen. It is suited for agentic visual workflows such as visual search, video captioning, and dense counting, and for physical-world use cases in robotics, construction, engineering, and manufacturing.

Qwen Image 2.1 Turbo is the fast, lower-cost tier of the Qwen Image 2.1 series from Qwen, for image generation and editing where throughput and price matter more than peak fidelity. See Qwen Image 2.1 Pro for the higher-quality tier.

Qwen Image 2.1 Pro is an image generation and editing model from Qwen, the higher-quality tier of the Qwen Image 2.1 series. Choose it over Qwen Image 2.1 Turbo when output fidelity matters more than cost or speed.

Decider Flash V1.2 is a 4B-parameter decision model from Perplexity, a smaller, lower-cost sibling of Decider V1.2 with the same API contract. Instead of generating text, it reads content passed as state (text, JSON, or images) and returns typed, probabilistic answers to one or more named questions in a single request: the probability of yes for a yes/no question (noul), a probability for every option plus the most likely one (choice, up to 255 options), or a probability for every level of an ordered rubric plus the expected score (score). It suits high-volume classification, routing, and moderation where application code thresholds the returned numbers rather than parsing a chat reply.

Decider V1.2 is a new 27B checkpoint of Perplexity's decision model, succeeding Decider V1.1 27B with the same API contract. Instead of generating text, it reads content passed as state (text, JSON, or images) and returns typed, probabilistic answers to one or more named questions in a single request: the probability of yes for a yes/no question (noul), a probability for every option plus the most likely one (choice, up to 255 options), or a probability for every level of an ordered rubric plus the expected score (score). It is built for classification, routing, moderation, and rubric grading where application code thresholds the returned numbers rather than parsing a chat reply. For a smaller, lower-cost option with the same contract, see Decider Flash V1.2.

Xing4.0-29B-A4B is an open-weight mixture-of-experts language model from China Telecom's AI team (TeleAI) and the successor to the TeleChat series. It has 29B total parameters with 4B active per token and a native 256K-token context window. It was trained entirely on Ascend NPUs with the MindSpore framework and is tuned for complex engineering and agent work: tool calling, multi-step planning, and long-context execution. Reasoning is on by default and can be turned off.

This model always redirects to the latest model in the Decider Flash family.

Beam 1.0 is Reflection's open-weight reasoning model for coding and long-horizon agentic workflows. It uses a sparse mixture-of-experts architecture with 501B total and 23B active parameters per token, and a 1M-token context window. Beam targets software engineering, multi-step tool use, and mathematical and scientific reasoning, with an emphasis on token efficiency. It takes text input and produces text output, supports configurable reasoning effort, function calling and structured outputs, and is released under Apache 2.0.

Mercury 3 is a diffusion-based language model from Inception, and the lab's most capable model to date. It is suited for agentic workflows and real-time applications where latency compounds, such as search agents, voice pipelines, and coding subagents. Its distinguishing trait is speed: more than 40x larger than the first Mercury model while generating at the same throughput, because it produces and refines tokens in parallel rather than one at a time.

Span-02 is an open-weight decision model from Respan. It reads a state of text and images and answers the questions you define about it: picking one of several options, a yes/no/unknown call, or a score on a scale. It is suited for evaluation, guardrails, routing, and monitoring of LLM and agent outputs at scale. It is the successor to Span-01.

d1-ultra is Liquid AI's largest decision model, served as a System One endpoint. Send a state of text and images along with typed questions, and it returns a choice, a score, or a yes/no answer, each with a probability taken directly from the model rather than written out as text. It uses the same /v1/systemone schema as other OpenRouter Decisions models, so it suits routing, classification, moderation, and visual checks that need a fast, scored answer instead of prose.

Krea K3 Edit is an image-to-image editing model from Krea. It takes one to eight reference images plus a text prompt and returns a single edited image, following instructions that range from small local changes to complex multi-step edits while keeping the subjects and composition of the references intact. The output follows the shape of the input unless an aspect ratio is requested. The default fast mode completes most edits in under ten seconds, which suits interactive editing tools and workflows that generate many variations of the same source image.

Moondream 3 Preview is a vision-language model from Moondream with 9B total parameters and 2B active (mixture-of-experts). It handles visual question answering, captioning, OCR, and visual reasoning over images, with a 32K-token context and an optional reasoning mode.

Moondream 3.1 is a vision-language model from Moondream, the successor to Moondream 3 Preview, with 9B total parameters and 2B active (mixture-of-experts). It handles visual question answering, captioning, OCR, and visual reasoning over images, with a 32K-token context and an optional reasoning mode.

Muse Spark Open is a coding-optimized open-weight model. Trained specifically for agentic coding workflows, it delivers higher first-attempt accuracy, reliable tool calling, and deep codebase understanding across the repos developers actually work in.