Funding

GLM Coding Plan starts at $18/month with 50% off-peak credits

Share
GLM Coding Plan starts at $18/month with 50% off-peak credits

Z.ai's GLM Coding Plan now starts at $18 per month for the Lite tier, billing users in credits with a 50% discount during off-peak hours—making it the only subscription among major open-weight coding models to meter usage this way.

TL;DR

  • GLM Coding Plan charges $18–$168/month in credits with 50% off-peak discounts, unlike per-token competitors like DeepSeek API and Mistral Large 4.
  • Kimi Code bundles a faster Highspeed tier but lacks documented off-peak pricing, while Mistral offers open weights for self-hosting with no subscription required.
  • Developers in the Americas benefit from off-peak discounts during their workday, while those in Europe face peak-hour pricing.

What happened

Z.ai's GLM Coding Plan introduced credit-based billing on July 30, 2026, replacing its previous token-based tiers. The plan now includes Lite ($18/month), Pro ($80/month), and Max ($168/month) tiers, with quarterly and yearly discounts of 20% and 30% respectively.

Each tier provides a five-hour rolling credit pool and a weekly cap, with Max offering 28,000 five-hour credits and 140,000 weekly credits. Off-peak usage (including weekends) consumes credits at half the standard rate, with peak hours defined as Monday to Friday from 14:00 to 18:00 Singapore time.

The plan exclusively supports GLM-5.3 and GLM-5.3-Flash models, with a 1M-token context window and up to 128K output tokens. Z.ai claims a 50% coding improvement for GLM-5.3 over its predecessor, though this is a vendor assertion.

Moonshot's Kimi Code offers K3.1 Pro with Highspeed variants, billed annually at 468–13,428 yuan (≈$39–$1,119/month) in China, with international pricing listed as $15–$159/month. DeepSeek API and Mistral Large 4 bill per token, with Mistral also providing open weights for self-hosting.

Why it matters

For developers, the GLM Coding Plan's credit system and off-peak discounts create predictable costs for bursty workloads, especially in the Americas where peak hours align with sleep time. European developers, however, face higher effective rates due to overlapping peak windows.

Teams with strict budgets may prefer GLM's subscription failure mode (credits exhausted = waiting period) over APIs that keep charging, while deadline-driven teams may favor APIs that fail open. The plan's restriction to a single account limits shared team access, pushing larger teams toward metered APIs like DeepSeek or Mistral.

Kimi Code's Highspeed tier appeals to users prioritizing latency, but its lack of transparent off-peak pricing and annual upfront billing adds friction. Mistral's open weights offer cost predictability for self-hosted setups but require infrastructure expertise.

The absence of token-to-credit conversion in GLM's documentation forces developers to estimate usage empirically, complicating budgeting for large-scale deployments.

Key facts

  • GLM Coding Plan Lite costs $18/month, Pro $80/month, and Max $168/month, with quarterly discounts of 20% and yearly discounts of 30%, according to PlanYield's November 2026 price check on z.ai/subscribe.
  • Max tier offers 28,000 five-hour credits and 140,000 weekly credits, priced at $1.20 per 1,000 weekly credits on the monthly plan, according to Z.ai's pricing page.
  • Off-peak usage (including weekends) consumes credits at 50% of the standard rate, with peak hours defined as Monday–Friday 14:00–18:00 Singapore time.
  • Kimi Code's K3.1 Pro membership tiers start at 468 yuan/year (≈$39/month) in China and $15/month internationally, with Highspeed variants available in Pro, Max, and Ultra tiers.
  • DeepSeek API bills per token with no minimum plan, while Mistral offers Large 4 at $0.70 per 1M input tokens and provides open weights for self-hosting under the Mistral Ecosystem Open-Weight license.
  • GLM Coding Plan exclusively supports GLM-5.3 and GLM-5.3-Flash models, with a 1M-token context window and up to 128K output tokens, according to Z.ai's FAQ and developer pages.
  • Kimi K3.1 Pro is described as Moonshot's most capable open-weight model at roughly 3 trillion parameters, with a 1M-token context window, according to the company's October 2026 announcement.
  • Mistral Large 4 weights were released for download and self-hosting on October 29, 2026, per the company's documentation.

Context

Open-weight coding models are reshaping developer workflows by offering transparency and cost control, with GLM, Kimi, DeepSeek, and Mistral representing distinct billing philosophies: subscription credits, annual memberships, pay-as-you-go tokens, and self-hosted weights.

The credit-based billing model used by GLM Coding Plan is rare among major coding model providers, aligning costs with usage patterns rather than raw token volume. This approach favors developers with intermittent, high-burst workloads over continuous, low-latency demands.

Annual upfront billing for Kimi Code and GLM's yearly discounts introduce cash-flow considerations for startups and bootstrapped teams, while Mistral's open weights cater to organizations with existing infrastructure for model deployment.

Topics

Join the discussion

Have a take on this story? Weigh in with our community on Facebook.

💬 Discuss on Facebook →