AI coding plans compared: price, limits, China vs global

No published message counts. Usage resets on a rolling 5-hour session window, with additional weekly limits on paid plans; web, desktop, mobile and Claude Code share one pool. Officially, Pro gives at least 5x Free's usage per 5-hour session. On Pro, Fable is available only via usage credits (API rates). Annual billing is $17/month ($200 billed up front).

No published message counts. 5x Pro's per-session usage (5-hour window), plus a weekly limit that resets at a fixed per-account time; Anthropic may add other weekly/monthly caps. Fable is capped at 50% of weekly limits. Includes $100/month in Claude Platform API credits. Monthly billing only; price is for web subscriptions (mobile may differ).

No published message counts. 20x Pro's per-session usage (5-hour window), plus a weekly limit; Anthropic may add other weekly/monthly caps. Fable is capped at 50% of weekly limits. Includes $200/month in Claude Platform API credits. Monthly billing only; price is for web subscriptions (mobile may differ).

OpenAI publishes ranges, not fixed caps, so limits are left empty. Estimated local messages per 5 hours — GPT-6 Astra 5–45, GPT-6.1 Sol 15–160, GPT-6 Sol 15–150, GPT-6 Luna 350–3,000; cloud tasks use more and weekly limits may apply. Codex and ChatGPT Work share usage; extra credits can be bought. GPT-5.5 retired from ChatGPT-sign-in Codex on 2026-10-14.

No published numbers; OpenAI only says Pro 200 includes more than Pro 100 and Pro 500 the most. Pro plans currently have no 5-hour limit; each model has its own allowance that resets. No Ultrafast. Monthly billing only.

No published numbers; Pro plans currently have no 5-hour limit. Pro 200 is open to new subscribers again, but with a lower allowance than before (price still $200). Users with an active Pro 200 between 2026-09-22 and 09-29 keep the old allowance until 2026-10-29, then move to the lower one. No Ultrafast. Monthly billing only.

No published numbers; highest allowance of the three Pro plans. Pro plans currently have no 5-hour limit. Only Pro tier with Ultrafast — for GPT-6 Astra and GPT-6.1 Sol it draws included usage at 8x Standard, then credits at 6x. Monthly billing only.

Credits are charged per token — (input×6.9 + cached×1.7 + output×24)/10000 for GLM-5.3 (Flash 2.3/0.56/8) — at 50% during off-peak hours (peak hours not defined on the page). Official estimate at 95% cache hit on GLM-5.3 is about 48–97M tokens/week. 5-hour credits roll; weekly credits reset every 7 days from purchase. Quarterly billing is 20% off and yearly 30% off. Only for supported coding tools, not general API use.

6x Lite usage. Credits are charged per token — (input×6.9 + cached×1.7 + output×24)/10000 for GLM-5.3 (Flash 2.3/0.56/8) — at 50% during off-peak hours (peak hours not defined on the page). Official estimate at 95% cache hit on GLM-5.3 is about 290–580M tokens/week. 5-hour credits roll; weekly credits reset every 7 days from purchase. Quarterly billing is 20% off and yearly 30% off. Only for supported coding tools, not general API use.

14x Lite usage. Credits are charged per token — (input×6.9 + cached×1.7 + output×24)/10000 for GLM-5.3 (Flash 2.3/0.56/8) — at 50% during off-peak hours (peak hours not defined on the page). Official estimate at 95% cache hit on GLM-5.3 is about 676–1,352M tokens/week. 5-hour credits roll; weekly credits reset every 7 days from purchase. Quarterly billing is 20% off and yearly 30% off. Only for supported coding tools, not general API use.

Same credit rules as the China edition — (input×6.9 + cached×1.7 + output×24)/10000 for GLM-5.3, 50% off-peak. Official estimate at 95% cache hit on GLM-5.3 is about 48–97M tokens/week. Quarterly 20% off, yearly 30% off.

6x Lite usage. Same credit rules as the China edition — (input×6.9 + cached×1.7 + output×24)/10000 for GLM-5.3, 50% off-peak. Official estimate at 95% cache hit on GLM-5.3 is about 290–580M tokens/week. Quarterly 20% off, yearly 30% off.

14x Lite usage. Same credit rules as the China edition — (input×6.9 + cached×1.7 + output×24)/10000 for GLM-5.3, 50% off-peak. Official estimate at 95% cache hit on GLM-5.3 is about 676–1,352M tokens/week. Quarterly 20% off, yearly 30% off.

No published numbers. New plans drop the weekly quota and keep only a rolling 5-hour rate window, plus a monthly total shared with the Kimi membership (frozen until reset or upgrade). Legacy tiers (Andante/Moderato/Allegretto…) keep the 7-day weekly quota. Plus gets K3 256K only; K3 1M context and HighSpeed need Pro or above. Annual billing $180/year. Extra Usage (pay-as-you-go, close to Open Platform API prices) is available.

No published numbers. New plans drop the weekly quota and keep only a rolling 5-hour rate window, plus a monthly total shared with the Kimi membership (frozen until reset or upgrade). Legacy tiers (Andante/Moderato/Allegretto…) keep the 7-day weekly quota. Includes K3 1M context (about 2x the quota of the 256K version) and HighSpeed. Annual billing $372/year. Extra Usage (pay-as-you-go, close to Open Platform API prices) is available.

No published numbers. New plans drop the weekly quota and keep only a rolling 5-hour rate window, plus a monthly total shared with the Kimi membership (frozen until reset or upgrade). Legacy tiers (Andante/Moderato/Allegretto…) keep the 7-day weekly quota. Includes K3 1M context (about 2x the quota of the 256K version) and HighSpeed. Annual billing $948/year. Extra Usage (pay-as-you-go, close to Open Platform API prices) is available.

No published numbers. New plans drop the weekly quota and keep only a rolling 5-hour rate window, plus a monthly total shared with the Kimi membership (frozen until reset or upgrade). Legacy tiers (Andante/Moderato/Allegretto…) keep the 7-day weekly quota. Includes K3 1M context (about 2x the quota of the 256K version) and HighSpeed. Annual billing $1908/year. Extra Usage (pay-as-you-go, close to Open Platform API prices) is available.

No published numbers, only multiples. Base tier. Text/image/audio are subject to a 5-hour and a weekly window (both start at first use). No video model. M Plan replaces the earlier Coding/Token Plan; legacy Token Plan subscribers keep their price and terms. Annual ¥490/year (pay 10 months).

No published numbers, only multiples. 3x Go usage. Text/image/audio are subject to a 5-hour and a weekly window (both start at first use). Includes the H3 video model (video is subject only to the weekly window). M Plan replaces the earlier Coding/Token Plan; legacy Token Plan subscribers keep their price and terms. Annual ¥1190/year (pay 10 months).

No published numbers, only multiples. 12x Go usage. China Build is 12x Go while global Build is only 7.5x. Text/image/audio are subject to a 5-hour and a weekly window (both start at first use). Includes the H3 video model (video is subject only to the weekly window). M Plan replaces the earlier Coding/Token Plan; legacy Token Plan subscribers keep their price and terms. Annual ¥4690/year (pay 10 months).

No published numbers, only multiples. Base tier. Text/image/audio are subject to a 5-hour and a weekly window (both start at first use). No video model. M Plan replaces the earlier Coding/Token Plan; legacy Token Plan subscribers keep their price and terms. Annual $220/year (pay 10 months).

No published numbers, only multiples. 3x Go usage. Text/image/audio are subject to a 5-hour and a weekly window (both start at first use). Includes the H3 video model (video is subject only to the weekly window). M Plan replaces the earlier Coding/Token Plan; legacy Token Plan subscribers keep their price and terms. Annual $550/year (pay 10 months).

No published numbers, only multiples. 7.5x Go usage. China Build is 12x Go while global Build is only 7.5x. Text/image/audio are subject to a 5-hour and a weekly window (both start at first use). Includes the H3 video model (video is subject only to the weekly window). M Plan replaces the earlier Coding/Token Plan; legacy Token Plan subscribers keep their price and terms. Annual $1320/year (pay 10 months).

Limited slots, first come first served; restocked daily at 09:30 (UTC+8). Lite closed to new orders on 2026-03-20 and to renewals/upgrades on 2026-04-13.

First month ¥39.90 for new customers, then ¥200/month. Quota counts model calls — about 5–10 per simple task and 10–30+ per complex one. The 5-hour quota rolls minute by minute; weekly resets Monday 00:00; monthly resets on the subscription date. Interactive use in coding tools only; no scripts or backends. Non-refundable.

Limited slots, first come first served; restocked daily at 00:00 (UTC+8). Lite closed to new orders on 2026-03-20 and to renewals/upgrades on 2026-04-13.

Same request quota as the China edition (counted per model call); $50 vs ¥200 in China. No first-month promo is stated. Interactive use in coding tools only. Non-refundable.

The 75%-off promo (2026-06-08 to 2026-11-08) has limited slots, first come first served, max two months per account.

No published numbers; only that 5-hour (from first request), weekly (Monday 00:00 reset) and monthly limits exist. List price ¥40/month; during the promo the first two months are ¥9.9 each, then list price. Quarterly/annual prices load dynamically and could not be confirmed. Coding tools only, no API use.

The 75%-off promo (2026-06-08 to 2026-11-08) has limited slots, first come first served, max two months per account.

No published numbers; officially 5x Lite, with 5-hour, weekly and monthly limits. List price ¥200/month; during the promo the first two months are ¥49.9 each, then list price. Quarterly/annual prices could not be confirmed. Coding tools only, no API use.