# Zhipu's GLM-5.3 Targets Coding at a Fraction of the Cost

> The open-weight MoE model from Zhipu AI aims to match frontier closed systems on coding tasks while undercutting them on price.

Published by The Open Weights on Aug 23, 2026. Canonical: https://theopenweights.com/news/glm-5-3-4zfa

## Key facts

- Company: Zhipu AI
- Model: GLM-5.3
- Version: 5.3
- Category: Text / LLM
- Modalities: Reasoning, Text / LLM, Code
- License: MIT (Open weights, commercial use allowed)
- Significance: notable
- Published: 2026-08-23
- Last verified: 2026-10-10
- Canonical URL: https://theopenweights.com/news/glm-5-3-4zfa

Zhipu AI has released GLM-5.3, an open-weight reasoning and coding model distributed under the permissive MIT license. The model uses a mixture-of-experts (MoE) architecture and is positioned as a serious challenger to frontier closed systems from labs like OpenAI and Anthropic — particularly on software engineering workloads.

The headline claim is economic as much as technical. According to a [Together.ai analysis](https://www.together.ai/blog/glm-5-3-vs-gpt-5-6-sol-on-deepswe-cost-coding-and-routing), GLM-5.3 competes with much pricier proprietary models on coding benchmarks while costing a fraction as much to run, with the write-up examining trade-offs across cost, coding performance, and routing.

## Why it matters

The gap between open and closed models on real coding tasks has been one of the last holdouts favoring proprietary providers. A capable open-weight model under MIT licensing changes the calculus for teams that want to self-host, fine-tune, or avoid per-token API bills.

- Open weights under the MIT license, allowing broad commercial use
- MoE design, which activates only part of the network per request to keep inference costs down
- Focus on reasoning and code, the areas where cost sensitivity is highest for developers

As with any vendor-adjacent benchmark, the cost and performance figures deserve independent verification, and Zhipu has not published full architectural specifications such as parameter counts or context length in this record. Still, for organizations weighing whether open models are ready for production coding, GLM-5.3 is a data point worth watching.

## Get the model

- [Hacker News](https://reinvently.co.uk/tools/ed-o-meter/)
- [Announcement](https://www.together.ai/blog/glm-5-3-vs-gpt-5-6-sol-on-deepswe-cost-coding-and-routing)
- [Hugging Face](https://huggingface.co/zai-org/GLM-5.3)

## Sources

- [GLM-5.3 (open-weight) beat Anthropic/OpenAI models – for 1/5 the cost](https://reinvently.co.uk/tools/ed-o-meter/) — Hacker News, Aug 23, 2026
- [GLM-5.3 vs. GPT-5.6 Sol on DeepSWE: Cost, Coding, and Routing](https://www.together.ai/blog/glm-5-3-vs-gpt-5-6-sol-on-deepswe-cost-coding-and-routing) — Announcement, Aug 21, 2026
- [zai-org/GLM-5.3](https://huggingface.co/zai-org/GLM-5.3) — Hugging Face, Aug 25, 2026

---
Source: The Open Weights (https://theopenweights.com/). Aggregated and written by Claude, curated by humans. Cite as: "Zhipu's GLM-5.3 Targets Coding at a Fraction of the Cost", The Open Weights, Aug 23, 2026, https://theopenweights.com/news/glm-5-3-4zfa