Open-weight frontier models with the refusal behavior removed, served without added content filters. Context, output limits, reasoning and the exact id to send, model by model.
| Model | Input | Context | Max output | Reasoning | Released |
|---|---|---|---|---|---|
DeepSeek V4.1 Flash 1017Newestdeepseek-v4.1-flash-1017-uncensored | TextImageText | 1M | 64K | Low → Max | Oct 2026 |
GLM 5.3glm-5.3-uncensored | TextText | 1M | 128K | Low → Max | Aug 2026 |
GLM 5.3 Flashglm-5.3-flash-uncensored | TextImageText | 1M | 32K | Low → Max | Aug 2026 |
Qwen 3.8 27Bqwen-3.8-27b-uncensored | TextImageText | 128K | 32K | On / off | Aug 2026 |
Kimi K3.1 Prokimi-k3.1-pro-uncensored | TextImageVideoText | 1M | 922K | Low → Max | Oct 2026 |
Hy4Coming soonhy4-uncensored | TextImageText | 1M | 64K | Low → Max | Nov 2026 |
Start with DeepSeek V4.1 Flash: it leads our models on reasoning, coding and security benchmarks. Use GLM 5.3 for long outputs or a second opinion, GLM 5.3 Flash for batch work, and Qwen 3.8 27B for quick chat.
The full guideThe newest and strongest of our models: ahead of the others on reasoning, coding and security benchmarks, and the fastest to answer.
Z.ai's coding and cybersecurity model, with the longest output we serve. Use it for long reports, large code changes and a second opinion from another model family.
A lighter GLM for batch jobs, labeling and triage, and a second model family when generated data should vary.
Our smallest model, served on our own deployment. Quick answers, chat and drafting, with thinking on or off.
Moonshot AI's most capable open-weight model, for agentic coding, research and very long outputs. It reads video as well as images and text.
Tencent's newest flagship, the full release after the Hy4 preview, with better long-context reading and longer agent runs. It takes images as well as text.
Tell us what you are building. We reply with an invitation, a key and pricing for your volume.