Z.ai Releases GLM-5.3 with 1M-Token Context and Always-On Reasoning
Z.ai has released GLM-5.3, a large-scale reasoning model aimed at software engineering and long-horizon agent tasks, featuring a 1M-token context window and mandatory reasoning that cannot be disabled. The model is priced at $1.40 per 1M input tokens and $4.40 per 1M output tokens on OpenRouter.
Z.ai has released GLM-5.3, a large-scale reasoning model built for complex software engineering and long-horizon agent tasks. The model supports a 1M-token context window for text input and output, and is available now through OpenRouter at $1.40 per 1M input tokens and $4.40 per 1M output tokens, with cached input priced at $0.26 per 1M tokens.
What's new
According to Z.ai, GLM-5.3 improves on its predecessor, GLM-5.2, in two areas: coding performance and the balance between output quality and token efficiency. The company has not published specific benchmark scores alongside the release, so these improvement claims remain unverified pending independent testing.
The model's reasoning process is always active and cannot be turned off — a departure from models that let users toggle chain-of-thought on or off. GLM-5.3 supports three reasoning effort levels: low, high, and max, with max set as the default. This effort setting likely trades off latency and token consumption against answer quality, though Z.ai has not disclosed exact performance deltas between the tiers.
The 1M-token context window positions GLM-5.3 for long-horizon agentic workloads — tasks that require the model to track extensive codebases, multi-step tool use, or extended conversation history without losing context.
Availability and reliability
OpenRouter data shows the model listed with a release date of August 18, 2026, and current uptime metrics indicate some rough edges: three-day uptime stands at 69.39%, with availability over the same period at 69.24%. Availability over the trailing 24 hours is lower still, at 36.16%. OpenRouter's routing infrastructure can fail over to alternate providers when the upstream Z.ai endpoint errors out, but users relying on this model directly should expect intermittent instability at this stage.
No throughput or latency figures were available at publication time, and OpenRouter notes there isn't yet enough usage data from apps built on the model to report broader adoption patterns.
What this means
GLM-5.3 continues Z.ai's push into the reasoning-model category, competing against offerings from OpenAI, Anthropic, DeepSeek, and Alibaba's Qwen line that also target coding and agentic use cases. The 1M-token context window is competitive with the upper end of what's currently available in the market, and the pricing — $1.40/$4.40 per 1M tokens — sits in a moderate range, cheaper than frontier proprietary models but not aggressively discounted.
The mandatory always-on reasoning is a notable design choice: it simplifies the API surface but removes flexibility for users who want fast, low-cost responses without a reasoning pass. Combined with the low uptime figures reported during this early window, teams evaluating GLM-5.3 for production coding agents should treat current benchmarks as provisional and monitor stability before committing critical workloads to the endpoint. Independent verification of Z.ai's coding and efficiency claims against GLM-5.2 has not yet appeared.
Related Articles
Qwen Launches Qwen3.8 27B, an Open-Weight Vision-Language Model with 262K Context
Qwen has released Qwen3.8 27B, a 27-billion-parameter dense vision-language model with a 262K token context window, available now via OpenRouter at $0.45 per million input tokens and $3.20 per million output tokens.
Z.ai Releases GLM-5.3, Claims Frontier Coding Scores From a 750B-Parameter Model
Z.ai released GLM-5.3, a coding-focused model built on the same base as GLM-5.2 but with substantially extended post-training, and claims it surpasses Moonshot AI's Kimi K3 on many agentic coding benchmarks despite having roughly a third of the parameters. The model is live in Z.ai's coding plan now, with API and open-weight Hugging Face access expected within two weeks.
Z.ai Releases GLM-5.3, Claims Frontier Agentic Coding Performance from 750B-Parameter Model via Post-Training Alone
Z.ai released GLM-5.3, available now in its coding plan, with API access and open weights on Hugging Face to follow within two weeks. The company says the model matches or beats larger frontier systems on agentic coding benchmarks using the same base checkpoint as GLM-5.2, with all gains coming from expanded post-training.
Zhipu AI Releases GLM-5.3, Claims It's the Strongest Open-Weights Coding Model
Zhipu AI has released GLM-5.3, a coding-focused model built on the same base as GLM-5.2 with additional post-training. The company claims it's the strongest open-weights coding model available, with gains concentrated in agentic and cybersecurity tasks, though independent benchmarks are not yet published.
Comments
Loading...