Z.ai (Zhipu) · Released Aug 14, 2026GLM-5.3
Flagship GLM model for long-horizon coding, agents, and complex project delivery.
- Capability
- 155.8ECI · #24 of 148
- Input
- $1.40per 1M tokens
- Output
- $4.40per 1M tokens
- Context
- 1M131K max output
What it costs
- Input
- $1.40per million tokens
- Output
- $4.40per million tokens
- Cached input
- $0.26per million tokens
- Blended (3:1)
- $2.15pricier than 72% of priced models
Typical monthly bills
| Monthly usage | Estimated cost |
|---|---|
| Side project2M input + 0.5M output tokens | $5.00 |
| Team assistant25M input + 5M output tokens | $57.00 |
| Production app250M input + 50M output tokens | $570.00 |
Official Z.AI API price, as listed on models.dev.
Independent benchmarks
#24 of 148 scored models
The shaded band is Epoch AI’s confidence range (153.7–158.3); the tick marks the median scored model.
Scores from Epoch AI, run independently of Z.ai (Zhipu).
The details
- Lab
- Z.ai (Zhipu)
- Released
- Aug 14, 2026
- Context window
- 1,000,000 tokens
- Max output
- 131,072 tokens
- Inputs
- Text
- Output
- Text
- Reasoning
- Adjustable effort low · high · max
- Tool calling
- Yes
- Structured output
- Yes
- Weights
- Open weights
- API model ID
glm-5.3on Z.AI- Availability
- 62 API providerslisted on models.dev
What Z.ai (Zhipu) claims
Published by the lab at launch. Settings vary, so compare these only with care.
| Benchmark | Score | Setting | Source |
|---|---|---|---|
| Terminal-Bench v2.1 | 88.2 | max effort | Source |
| Terminal-Bench v3.0 | 28.3 avg@3 | max effort | Source |
| DeepSWE v1.1 | 66.9 resolved | max effort; 6h timeout; 400K context | Source |
| NL2Repo | 58 | max effort | Source |
| ProgramBench | 19 almost solved | max effort | Source |
| FrontierSWE | 78.1 | max effort | Source |
| SWE-Marathon v1.1 | 42.5 | max effort; modified anti-cheat checks | Source |
| PostTrainBench | 39.8 weighted average | max effort; 3 runs; modified external-API checks | Source |
| CyberGym | 84.5 | max effort | Source |
| ExploitGym | 105 tasks solved | max effort; 2h rescaled timeout | Source |
| ExploitGym | 130 tasks solved | max effort; 6h rescaled timeout | Source |
| ExploitBench | 54.4 average coverage | max effort; union over 3 revisions | Source |
More from Z.ai (Zhipu)
About GLM-5.3
How much does GLM-5.3 cost?
GLM-5.3 costs $1.40 input / $4.40 output per million tokens (official Z.AI API price). Cached input is $0.26 per million tokens. At a 3:1 input-to-output mix that is $2.15 per million tokens, more expensive than 72% of the 360 priced models we track.
What is the context window of GLM-5.3?
GLM-5.3 accepts up to 1,000,000 tokens per request and can write up to 131,072 tokens in one response.
How good is GLM-5.3?
Epoch AI gives GLM-5.3 a Capabilities Index score of 155.8 (likely range 153.7–158.3), ranking it #24 of 148 models Epoch has scored. Epoch AI benchmark results: GPQA Diamond 90.9%, FrontierMath Tiers 1–3 68.8%, OTIS Mock AIME 2024–2025 91.1%, SimpleQA Verified 41.0%.
Is GLM-5.3 open source?
Yes. Z.ai (Zhipu) publishes the weights, so it can be downloaded and self-hosted.
What inputs does GLM-5.3 support?
GLM-5.3 accepts text and replies in text. It is a reasoning model with low, high and max effort settings, supports tool calling and can return structured JSON output.
When was GLM-5.3 released?
Z.ai (Zhipu) released GLM-5.3 on Aug 14, 2026.
What are the best alternatives to GLM-5.3?
The closest current models from other labs on capability, price and release date are Muse Spark 1.2 (Meta, ECI 155.0, $1.25 / $4.25), Grok 4.6 (xAI, ECI 156.6, $2.00 / $6.00), Qwen3.8 Max 0902 (Alibaba (Qwen), ECI 155.2, $2.00 / $6.00) and Gemini 3.8 Flash (Google, ECI 156.9, $0.75 / $3.75).