Chinese AI Models Compared on magicdoor.ai

Compare current GLM, DeepSeek, Kimi and Qwen model options by task, price and input support; MiniMax M3 is historical.

Kimi update — October 28, 2026: Choose Kimi K3.1 Pro for the current Kimi route through Moonshot AI via Vercel AI Gateway. K3 and Renewable K3 are retired. See the Kimi workflow and limits guide before continuing an older Kimi chat.

Direct answer: Start with GLM-5.3 Flash for low-cost iterations, DeepSeek V4.1 Flash 1017 for a text task, or Kimi K3.1 Pro for a different multimodal or coding perspective. MiniMax M3 was a previous option and is no longer in the picker. Check the live model directory before assuming availability.

DeepSeek checkpoint update — October 21, 2026: Current DeepSeek labels, links and any DeepSeek prices on this page now refer to V4.1 Flash 1017, using the October 21 launch source packet. Earlier fact-check and source-access dates below apply to the original comparison or study, not this checkpoint. Other vendor claims and historical measurements were not reverified in this update.

Original comparison reviewed October 8, 2026 against the planned Magicdoor registry migration. The DeepSeek row was updated October 21 using the checkpoint sources linked above; other rows were not reverified. Rates are standard per-million input/output tokens and exclude cache qualifications.

ModelStandard input / outputTask starting point
GLM-5.3 Flash$0.15 / $0.50Cheap short iterations, including image-input comparisons
DeepSeek V4.1 Flash 1017$0.30 / $1.20Text-first analysis on the pinned zero-data-retention route
GLM-5.3$1.40 / $4.40Longer text reasoning
Qwen3.8 Max$2.50 / $6.25A second perspective for demanding text work
Kimi K3.1 Pro$2 / $10Coding or multimodal comparison

These are starting points, not a benchmark ranking. The same prompt, source context, acceptance criteria and review process make a fairer test than a model name or token rate alone. Cached tokens may be priced differently; see cost per model. You can start a chat and switch models mid-conversation when the selected model supports the conversation and its attachments.

FAQ

Is MiniMax M3 still in the current picker?

No. MiniMax M3 is retired from new model selection. Previous chat history is preserved through the runtime compatibility layer.

Which low-cost model should I try first?

Try GLM-5.3 Flash at $0.15 input and $0.50 output per million tokens, or DeepSeek V4.1 Flash 1017 at $0.30 and $1.20. Compare an actual task before judging quality or total cost.