AI Models · Pricing

GLM-5.3-Flash Pricing: API Cost, Launch Discount & Real-World Usage

Pricing captured August 30, 2026 · ToolStep Editorial Team

GLM-5.3-Flash API pricing as of August 30, 2026: $0.15 per million input tokens, $0.03 per million cached input tokens, $0.50 per million output tokens, with a 50% launch discount valid until September 9, 2026 24:00 (UTC+8). Zhipu prices the model at 1/10 of its GLM-5.3 flagship (1/20 during the discount) and about 1/40 of Claude Opus 4.8 — which makes it the cheapest 1M-context, coding-capable endpoint in the current open-model cohort. This page breaks down the official rates, what they mean at real usage levels, and how they compare with the models that shipped in the same week.

Pricing snapshot

Captured: August 30, 2026. Attribution note: Zhipu's official docs state only relative pricing (1/10 of GLM-5.3; 1/20 during the launch discount; ~1/40 of Claude Opus 4.8). The absolute USD figures below are as reported by third-party fact sheets (list tier) and as listed on OpenRouter, checked on 2026-08-30 (discount tier — OpenRouter currently serves the model at $0.075/$0.25/$0.015). The launch discount expires September 9, 2026 24:00 (UTC+8); list prices apply afterward. Re-verify in your billing console before budgeting — vendor pricing pages change without notice.

Official GLM-5.3-Flash API Pricing

Token typeList price (per 1M tokens)Launch price (50% off, until Sept 9, 2026 24:00 UTC+8)
Input$0.15$0.075
Cached input$0.03$0.015
Output$0.50$0.25
Cache storageTemporarily free

Attribution: the launch-discount tier ($0.075 / $0.015 / $0.25) is confirmed by the OpenRouter listing (checked 2026-08-30); the list tier ($0.15 / $0.03 / $0.50) is as reported by third-party fact sheets reproducing Zhipu's published schedule. Zhipu's official docs state the relative framing only: GLM-5.3-Flash costs 1/10 of GLM-5.3 at list (1/20 during the launch discount) and about 1/40 of Claude Opus 4.8. For comparison, DeepSeek V4 Pro (pre-August 16 schedule) was $0.435 input / $0.87 output — GLM-5.3-Flash undercuts even that on input by ~3x at list. Note the flagship GLM-5.3 still has no published standalone API price as of this capture; it is available via the GLM Coding Plan's points system only (see our GLM-5.3 review).

Region & Channel Notes

GLM Coding Plan: Quota & Off-Peak Rates

For coding-agent usage, GLM-5.3-Flash is bundled into the GLM Coding Plan with a 3x quota compared to GLM-5.3, and calls outside peak hours consume 50% of standard points. Peak hours are weekdays 14:00–18:00 (UTC+8); weekends are entirely off-peak. Input, cached input, and output tokens are metered separately under the plan's points system, and ZCode adds a 1.5x limited-time quota boost through August 31, 2026. If you live in an IDE/agent workflow rather than raw API calls, the Coding Plan is usually the cheaper route than pay-per-token — the opposite of the recommendation for pure API services.

Real-World Cost Math

Monthly token volumes converted to dollars at GLM-5.3-Flash rates (list / launch-discounted):

Usage profileMonthly tokens (in / out)Cost at listCost during 50% discount
Light API user10M in / 2M out$2.50$1.25
Typical dev tool50M in / 15M out$15.00$7.50
Heavy agent workload300M in / 100M out$95.00$47.50
High-cache RAG app (80% cache hits)300M in / 20M out, 240M cached240M × $0.03 + 60M × $0.15 + 20M × $0.50 = $24.70$12.35

Three practical takeaways. First, output tokens dominate — the 3.3x input/output price ratio means agentic workloads (which emit lots of code and tool calls) should be quoted on output volume, not input. Second, caching is the lever that matters: at $0.03/M, cached input is 5x cheaper than fresh input, so systems with stable system prompts (RAG, agents with persistent instructions) get an outsized benefit — and cache storage is temporarily free. Third, the launch discount effectively halves every row above until September 9.

GLM-5.3-Flash vs. Same-Week Competitors

The August 26–28 window produced three open-model price points to compare directly. Qwen3.8-Flash prices are published in CNY (¥0.8 input / ¥2.7 output / ¥0.1 cached per million); Hy4 preview at ¥6 / ¥18 with cache hits from ¥0.3. Approximate USD conversions (at ~7.1 CNY/USD, marked as estimates):

Model (as of Aug 30, 2026)Input / 1MOutput / 1MCached input / 1MContext
GLM-5.3-Flash (list)$0.15$0.50$0.031M
GLM-5.3-Flash (launch, until Sept 9)$0.075$0.25$0.0151M
Qwen3.8-Flash (API)¥0.8 (≈$0.11)¥2.7 (≈$0.38)¥0.1 (≈$0.014)1M
Tencent Hy4 preview (API)¥6 (≈$0.85)¥18 (≈$2.54)from ¥0.3 (≈$0.04)1M+
DeepSeek V4 Pro$0.435$0.87 → $3.96 peak / $1.98 off-peak$0.0036251M
GPT-5.6 Luna (OpenAI budget tier)$0.20$1.20

Read more in our GLM-5.3-Flash vs Qwen3.8-Flash-Next and price & speed comparison, plus the Hy4 preview review and DeepSeek V4 Pro review for their pricing details.

Who Is GLM-5.3-Flash Pricing For?

Things to Watch

Bottom Line

At $0.15 / $0.50 per million tokens — half that until September 9 — GLM-5.3-Flash is the value anchor of the August open-model cohort: the cheapest 1M-context coding-capable API, with a 5x-cheaper cache tier and free cache storage on top. The price is only half the story, though: its review page documents where the efficiency trade-offs show up (reasoning, some vision suites). Our independent take: for budget-constrained agent and RAG builders, this is the default starting point this week — but budget on list prices, not the discount, for anything past mid-September.

FAQ

What is GLM-5.3-Flash API pricing?

$0.15 / $0.03 / $0.50 per million tokens (input / cached input / output) at list, captured August 30, 2026; cache storage temporarily free.

How long does the launch discount last?

The 50% discount ($0.075 / $0.015 / $0.25) is valid until September 9, 2026 24:00 (UTC+8).

Is GLM-5.3-Flash cheaper than GLM-5.3?

Yes — 1/10 of GLM-5.3's price per Zhipu (1/20 during the discount). GLM-5.3's standalone API price was unpublished as of August 30, 2026.

Does the GLM Coding Plan include GLM-5.3-Flash?

Yes — fully available with 3x the quota of GLM-5.3, and 50% point consumption outside peak hours (weekdays 14:00–18:00 UTC+8, plus weekends).

How much does GLM-5.3-Flash cost at scale?

50M in / 15M out per month ≈ $15 at list ($7.50 discounted). Heavy agent use (300M in / 100M out) ≈ $95 at list ($47.50 discounted). Cache-heavy RAG cuts input cost 5x.

Sources (accessed August 30, 2026): Zhipu official docs (docs.bigmodel.cn — GLM-5.3-Flash relative pricing), z.ai/subscribe (GLM Coding Plan), OpenRouter listing for z-ai/glm-5.3-flash (discount-tier prices, checked 2026-08-30), third-party fact sheets reproducing Zhipu's published price schedule (list tier, discount deadline), DeepSeek V4 Pro official pricing as documented in our DeepSeek V4 Pro review, Qwen3.8-Flash pricing from Alibaba's official platform and OpenRouter listing, Hy4 pricing from Tencent's official announcement. USD conversions for CNY prices are estimates at ~7.1 CNY/USD. ToolStep claims no laboratory test results; prices re-verify in vendor consoles.