Mercury 3

Mercury 2.5 Preview

Mercury 2.5 Preview is Inception's latest and most intelligent diffusion language model. Instead of generating tokens strictly one at a time, it produces and refines multiple tokens in parallel, reaching up to 1,107 tokens per second on standard GPUs. It delivers a 10+ point intelligence gain over Mercury 2, with tunable reasoning, parallel tool calls, schema-aligned JSON output, and a 260K context window. It is built for latency-sensitive production work such as search agents, voice pipelines, customer support, rapid coding iteration, and coding subagents.

Mercury 2

Inception Labs' fastest reasoning model with tool calling and structured outputs support.

Mercury Coder Small

Model by Inception AI. A diffusion large language model that runs incredibly quickly (500+ tokens/second) while matching Claude 3.5 Haiku and GPT-4o-mini. 1st in speed on Copilot arena, and matching 2nd in quality.

MiniMax M3.1 Flash

MiniMax M3.1 Flash is a multimodal model for coding, tool use, and long-context work. It accepts text, images, and video and produces text, with a 1,000,000-token context window and up to 524,288 output tokens within the remaining context. Thinking is always enabled, with low, medium, high, xhigh, and max effort levels; max is the default. Thinking tokens count toward the output limit.

DeepSeek V4.1 Flash 1017

The October 17, 2026 snapshot of DeepSeek V4.1 Flash supports text and image input, tool calling, and optional reasoning with low, high, and max effort. It has a shared 1,048,576-token context window and supports up to 393,216 output tokens within the remaining context. Zero data retention is available on eligible routes; output limits and reasoning levels depend on the selected route. Released under the MIT license.

DeepSeek V4.1 Flash 1017 Thinking

The October 17, 2026 snapshot of DeepSeek V4.1 Flash supports text and image input, tool calling, and optional reasoning with low, high, and max effort. It has a shared 1,048,576-token context window and supports up to 393,216 output tokens within the remaining context. Zero data retention is available on eligible routes; output limits and reasoning levels depend on the selected route. Released under the MIT license.