GLM-5.3-FLASH · SVG
A self-portrait in jade.
A glossy green bull with ivory horns, gold hooves and a lifted front leg.
Benchmark task
A rounded jade bull raising a hoof tests anatomy, balance and the appearance of polished stone and metal.
Revealed · Z.ai GLM-5.3-Flash
Ox Alpha,
revealed. GLM-5.3-Flash
Multimodal coding, long agent tasks, and open weights. Meet Z.ai’s GLM-5.3-Flash.
Explore the modelVISION
Images + videoGive it screenshots, diagrams or screen recordings. Native visual input helps it understand interfaces and use visual feedback while coding.
Z.ai · visual capabilities ↗1Mcontext · Z.ai
18B / 320Bactive / total parameters
MITopen-weight license
Tap the ox to stomp and explore.
Advertisement Demodokos Foundry
A whole radio station.
One graphics card.
Everything you hear on Ox FM was generated in Demodokos Foundry on a home PC: the music, the singing and every word the DJs speak. No credits, no queue, nothing uploaded.
- Songs with vocals
- Speech with 40+ emotions
- Windows, 6 GB+ GPU
Full access. Cancel anytime.
Ox FM
Five stations, one home PC
88.1 FMNo Permission
On airFoundry speech · EmmaNow playing the song
You are listening to Ox FM. Every song on this station was made on a home graphics card. No credits. No queue. And definitely, no permission.
Your station is next. Make a song tonight
Music and speech, both created with Demodokos Foundry
Audio could not load. Hear the songs at Demodokos ↗
OX ALPHA · STEALTHMARK
What Ox Alpha can draw.
Four StealthMark benchmark tasks from GLM-5.3-Flash. Explore its SVG results and benchmark tasks.
GLM-5.3-FLASH · SVG
Mona Lisa, in vectors.
A framed portrait with folded hands and a hazy mountain landscape.
Benchmark task
An SVG interpretation of the Mona Lisa tests facial likeness, anatomy and the composition of a familiar portrait.
GLM-5.3-FLASH · SVG
The wings are taking a break.
A helmeted pelican takes a red bicycle along the seafront.
Benchmark task
A cycling pelican tests anatomy, riding posture and contact between the bird and the bicycle.
GLM-5.3-FLASH · SVG
The beach is still busy.
Palms, reflections, animals and a flying kite fill a colourful beach scene.
Benchmark task
A busy shoreline at sunset tests scene composition, animal forms and the relationship between light and water reflections.
· GLM-5.3-Flash.
OX ALPHA · BENCHMARKS
How Ox Alpha
compares.
The released model, GLM-5.3-Flash, appears alongside Qwen 3.8, GPT, Claude and Gemini.
Ox Alpha / GLM-5.3-FlashIndependent results · Artificial Analysis27 September 2026
GRADUATE-LEVEL SCIENCE
GPQA Diamond
91.2%
198 questions in physics, chemistry and biology. Measures scientific knowledge and reasoning.
9 models · highest score first · scroll to compare ↕
HLE · TEXT-ONLY · NO TOOLS
Humanity’s
Last Exam
39.9%
2,158 expert-level questions across mathematics, humanities and science.
9 models · highest score first · scroll to compare ↕
91.2% on GPQA Diamond. 39.9% on HLE. GLM-5.3-Flash scores above Qwen3.8-Flash-Next on HLE; Qwen 3.8 Max leads it on both reasoning tests.
CODING & LONG CONTEXT
Code tasks and long document reasoning.
Higher is better · gold underline = best shown
Sorted by Terminal-Bench 2.1 · highest first
Swipe the table to compare all three tests. Model names stay in view.
| Model / reasoning | Terminal-Bench 2.1Complete terminal tasks | SciCodeScientific Python · subproblems | AA-LCR v1.1Reason across long documents |
|---|---|---|---|
| Qwen 3.8 Max0902 · reasoning | 88.8% | 52.1% | 80.3% |
| GPT-6 Astramax | 88.4% | 56.5% | 80.7% |
| GPT-5.6 Solmax | 88.0% | 57.1% | 84.0% |
| GPT-5.6 Terramax | 88.0% | 55.0% | 83.0% |
| Qwen3.8-Flash-NextReasoning | 86.1% | 50.6% | 79.7% |
| Gemini 3.7 Flashhigh | 85.8% | 57.2% | 81.7% |
| Claude Opus 4.8Adaptive · max | 84.6% | 54.4% | 77.7% |
| GLM-5.3-FlashOx Alpha · max | 84.3% | 51.6% | 80.0% |
| Qwen3.8-27Bxhigh | 79.8% | 46.6% | 82.0% |
Test settings & sources
- Reasoning tests
- GPQA uses the 198-question Diamond subset. HLE uses the 2,158 text-only questions from the May 2025 revision, without tools. Both report pass@1.
- Coding tests
- Terminal-Bench 2.1 uses Terminus 2 across 89 tasks, averaged over three runs. SciCode scores 288 Python subproblems with scientific background provided, also over three runs.
- Long context
- AA-LCR v1.1 uses 100 questions over documents of roughly 100K tokens, averaged over three runs. Model links open the evaluator’s individual result pages.
Reasoning effort is shown beside each model; Qwen 3.8 Max is the 0902 release. Scores are rounded to one decimal. Artificial Analysis methodology ↗
Z.AI RELEASE EVALUATION
Code & automation
Score / 100Sorted by DeepSWE · highest first
| Model | DeepSWEv1.1 | Automation |
|---|---|---|
| GPT-5.6 Terra | 69.6 | 37.2 |
| Gemini 3.7 Flash | 65.3 | 52.3 |
| GLM-5.3-Flash | 63.4 | 48.8 |
| DeepSeek-V4-Vision-Exp | 59.3 | 38.8 |
| Claude Opus 4.8 | 58.0 | 41.0 |
| GLM-5.2 | 46.2 | 26.2 |
DeepSWE tests software engineering; AutomationBench tests automated workflows. Z.ai’s DeepSWE run uses mini-swe-agent, 400K context and a 6-hour timeout. Release chart ↗ · Evaluation settings ↗
6 models · scroll to compare ↕
Z.AI API · USD / 1M TOKENS
What the released model costs.
- Input
- $0.15
- Cached input
- $0.03
- Output
- $0.50
Current standard rates. The free Ox Alpha preview has ended.
Official pricing ↗Benchmarks checked 27 September 2026 · Read the identity reveal ↗
ONE MODEL · TWO NAMES
Ox Alpha is GLM-5.3-Flash.
From ox-alpha to glm-5.3-flash
Z.ai previewed GLM-5.3-Flash under the codename Ox Alpha, using stealth/ox-alpha on OpenRouter. The released model is also written as GLM 5.3 Flash. Read Z.ai’s model overview ↗
Use the released model
The Z.ai API uses glm-5.3-flash; OpenRouter uses z-ai/glm-5.3-flash. For Z.ai, create an account and API key, then add a token balance as needed. The weights are available under the MIT license.