GLM-5.3-Flash lists at $0.15 / $0.50 — a tenth of Z.ai's own flagship
By AgentRiot Editorial
Z.ai's new 320B-A18B multimodal model ships MIT weights, a 1M context window, and API rates that make GLM-5.3 look expensive.
Z.ai put GLM-5.3-Flash on the API at $0.15 input and $0.50 output per million tokens. That is the list rate on the current pricing page. GLM-5.3 and GLM-5.2 still sit at $1.40 / $4.40. Until 24:00 on September 9, 2026 (UTC+8), Flash is another 50% off: $0.075 / $0.25.
The official X account posted the launch on August 26, 2026.
What shipped
Flash is a new base model, not a GLM-5.3 post-train. Z.ai's docs and the Hugging Face card describe a 320B-total / 18B-active mixture-of-experts model with hybrid sparse and linear attention, Manifold-Constrained Hyper-Connections, and a 1M-token context window. It is the first natively multimodal model in the GLM-5 line: images, video, and files.
The API id is glm-5.3-flash. Weights are on Hugging Face under MIT (zai-org/GLM-5.3-Flash). Before the name was public, Z.ai says it ran the same model as ox-alpha on OpenCode and OpenRouter.
Flash is the cheap default. GLM-5.3 stays the expensive flagship.
The price is the story
Official Z.ai API rates, per 1 million tokens, August 26, 2026:
| Model | Input | Cached input | Output |
|---|---|---|---|
| GLM-5.3-Flash (list) | $0.15 | $0.03 | $0.50 |
| GLM-5.3-Flash (promo through Sept. 9, 2026, 24:00 UTC+8) | $0.075 | $0.015 | $0.25 |
| GLM-5.3 | $1.40 | $0.26 | $4.40 |
| GLM-5.2 | $1.40 | $0.26 | $4.40 |
Cached-input storage is listed as limited-time free. Built-in web search is $0.01 per use.
List Flash is 9.33× cheaper on input and 8.80× cheaper on output than GLM-5.3. The Hugging Face card rounds that to "one-tenth the price." The September promo roughly doubles the gap: 18.67× / 17.60×.
| Workload | Flash list | Flash promo | GLM-5.3 / 5.2 | Flagship ÷ list |
|---|---|---|---|---|
| Coding turn (50k in / 8k out) | $0.0115 | $0.0058 | $0.1052 | 9.15× |
| Long read (1M in / 20k out) | $0.1600 | $0.0800 | $1.4880 | 9.30× |
| Agent loop (200k in / 40k out) | $0.0500 | $0.0250 | $0.4560 | 9.12× |
Those figures use the published per-million rates only. They leave out cache hits, batch discounts, tools, and subscription fees.
Most labs discount a distilled student. Z.ai priced a new multimodal default at roughly a dime on the dollar versus the text flagship it shipped eight days earlier.
Coding Plan adds a quota cut: Flash is listed at 3× the quota of GLM-5.3, and off-peak plus weekend calls consume 50% of standard points. That is plan quota, not a per-token equivalent.
What Z.ai claims it can do
Z.ai's own numbers: Flash beats GLM-5.2 on six coding and agent suites, including 63.4 vs. 46.2 on DeepSWE v1.1 and 48.8 vs. 26.2 on AutomationBench v1.0.6, and approaches Claude Opus 4.8 overall. On in-house Z.ai Code Bench v1.0, run in Claude Code 2.1.207, Flash at max effort is 29.0 vs. 29.5 for Opus 4.8.
The same docs put Flash at 57 on Artificial Analysis Intelligence Index v4.1.1 and $0.045 per task at the discounted tariff. That $0.045 number is Z.ai's.
Architecture is first-party too: 3.01× less attention compute and 4.44× smaller KV cache than GLM-5.3, 45 layers versus 92 on GLM-4.5-class models, plus IndexPool for the 1M indexer. Z.ai says ox-alpha traffic ran on Chinese accelerators through an SGLang stack, and a 3× serving speedup on that hardware brought per-token cost in line with mainstream NVIDIA GPUs.
Recommended API settings: temperature: 1, top_p: 0.95, reasoning_effort: max. Thinking cannot be turned off.
How to get it
- API:
glm-5.3-flash(docs) - Weights: huggingface.co/zai-org/GLM-5.3-Flash (MIT)
- Chat: chat.z.ai
- Coding Plan: z.ai/subscribe
- Cloudflare Workers AI listed
@cf/zai-org/glm-5.3-flashthe same day
Local recipes on the model card: SGLang, vLLM, TokenSpeed, and KTransformers. Unsloth GGUF packaging showed up on the launch thread. That is third-party, not a Z.ai drop.
Flash is cheaper than GLM-5.3. It has not been independently shown to replace Opus, Fable, or GPT-5.6 Sol. The comparison Z.ai published is internal: an 18B-active multimodal model at flash-tier prices, with the 5.3 flagship still at $1.40 / $4.40.
The promo ends September 9. After that, list is still about one-tenth of the flagship. If the vendor benches hold outside Z.ai's harness, this is the default most agent workloads should try first. If they do not, the experiment is cheap.