Plan recommendations

Why we recommend it：At ¥49 per month, it includes roughly 600M+ M3 tokens per month, support for three to four concurrent agents, a 1M-token context window, and multimodal usage. Text, image, voice, and music share the same quota. M3 is not a leading model, but it is a solid second-tier option for everyday work.

Note：The plan mainly covers MiniMax's own models, so it is not ideal if you frequently switch between model providers. Start with Plus; upgrade to the ¥119 Max plan only if you actually need more quota. New users generally do not need Ultra.

Why we recommend it：It costs $5 for the first month and $10 per month afterward, with access to GLM, Kimi (including K3), MiniMax, DeepSeek, and other models. It is a good fit if you have not settled on a preferred model yet. Usage is also relatively transparent: OpenCode publishes five-hour, weekly, and monthly request limits for each model.

Note：WeChat Pay and Alipay are supported. Quota is calculated according to model cost, so more expensive models consume it faster. Go has a single tier and is suitable for light to moderate use.

Why we recommend it：The $20-per-month Pro plan includes at least $20 of API usage across GPT, Claude, Gemini, and many other models. That allowance alone is limited, but Cursor also provides a first-party usage pool for Composer 2.5 and Grok 4.5. Both models are tuned by Cursor and perform well for coding inside Cursor. This pool currently provides roughly $300–$400 worth of usage per month, making the plan good value.

Details：Alipay is supported. The plan can only be used inside Cursor; its quota cannot be used in other agents such as Codex or Claude Code.

Why we recommend it：From June 8 through November 8, 2026, all users can receive up to 75% off the first two months of Coding Plan Lite or Pro. Lite is ¥9.9/month instead of ¥40, and Pro is ¥49.9/month instead of ¥200; regular pricing resumes in the third month. The official Coding Plan currently supports doubao-seed-2.1-turbo, doubao-seed-2.0-lite, MiniMax M2.7, MiniMax M3, GLM-5.2 (official identifier: glm-latest), DeepSeek V4 Flash, DeepSeek V4 Pro, Kimi K2.6, and Kimi K2.7 Code.

Note：Availability is limited and first-come, first-served. Each account receives at most two discounted months, shared across new purchases, renewals, and upgrades; changing tiers or cancelling does not grant a new benefit. The separate 5%-off referral promotion can be combined. Kimi K3 is not included in Coding Plan. See the official Volcengine promotion rules and checkout page for final eligibility.

Why we recommend it：There is no purchase queue, and the plan supports GLM-5.2. Lite, Pro, and Max cost approximately $18, $72, and $160 per month. An invitation link provides an additional 10% discount.

Note：It costs more than the domestic version. Start with Lite for light use and upgrade only after you confirm that the quota is insufficient. Do not buy a higher tier solely to avoid waiting for the domestic plan.