GLM-5.3
Z.ai · released
Z.ai's text-only flagship with a 1M-token context and downloadable weights, aimed at coding agents and long tasks.
Best for: Open-weight coding agents · Long-context text work · Self-hosting at scale
- Intelligence Index
- 44.8
- #11 of 38 · at Max effort
- Price per 1M tokens
- $1.4 in $4.4 out
- Blended $2.15 at 3:1
- Context window
- 1M
- Up to 128K output
- Output speed
- 79 tokens/s
- $2.01 measured cost per task
Capability profile
One 0–100 score per area, placing this model between the weakest and strongest tracked model across that area's benchmarks.
- Overall71 · #9 of 38
Based on 3 of 3 benchmarks
- Coding68 · #8 of 39
Based on 5 of 5 benchmarks
- Agents & knowledge work73 · #7 of 40
Based on 6 of 6 benchmarks
- Reasoning & science62 · #15 of 39
Based on 2 of 2 benchmarks
- Knowledge & reliability71 · #15 of 39
Based on 2 of 2 benchmarks
- VisionNot enough data
Based on 0 of 2 benchmarks
Reasoning effort trade-off
The same model at each effort setting. Lower settings give up a few points for a fraction of the cost and wait.
| Setting | Intelligence Index | Cost / task | First answer |
|---|---|---|---|
| Max | 44.8 | $2.01 | 28s |
| Low | 34.3 | $0.85 | 28s |
Measured by Artificial Analysis. First answer is the median wait until the first answer token, including thinking time.
Benchmark results
All third-party results on record. Rank is among the tracked models that have been tested on that benchmark.
Overall
#11 of 38· Max effort· Artificial Analysis
- Vals Index53.5%
#12 of 26· Max effort· $7.25 per test· Vals AI
- Arena · Text1478 ±6
#11 of 33· Max effort· 17,857 votes· Arena
Coding
- Terminal-Bench 4.041.9%
#7 of 38· Max effort· Artificial Analysis
- SciCode59.0%
#7 of 38· Max effort· Artificial Analysis
- Arena · WebDev1622 ±8
#15 of 34· Max effort· 8,195 votes· Arena
- Vibe Code Bench78.1%
#18 of 26· Max effort· Vals AI
- Code Migration43.2%
#13 of 26· Max effort· Vals AI
Agents & knowledge work
- GDPval-AA57.6
#8 of 39· Max effort· Artificial Analysis
- Arena · Agent+2.4 ±0.7
#15 of 29· Max effort· Arena
- Finance Agent v255.8%
#9 of 26· Max effort· Vals AI
#20 of 26· Max effort· Vals AI
- Legal Research Bench49.0%
#5 of 26· Max effort· Vals AI
- Tax Agent Bench73.1%
#4 of 26· Max effort· Vals AI
Reasoning & science
- Humanity's Last Exam42.3%
#17 of 39· Max effort· Artificial Analysis
- CritPt19.1%
#13 of 39· Max effort· Artificial Analysis
- ARC-AGI-2saturatingNot tested
- ARC-AGI-3Not tested
- GPQA Diamondsaturating91.7%
#12 of 27· Max effort· Artificial Analysis
Knowledge & reliability
#13 of 39· Max effort· Artificial Analysis
- AA-LCR79.7%
#23 of 39· Max effort· Artificial Analysis
Vision
- MMMU-ProNot tested
- Arena · VisionNot tested
Specifications
- Lab
- Z.ai
- Released
- Aug 18, 2026
- Availability
- Generally available
- Context window
- 1,000,000 tokens
- Max output
- 128,000 tokens
- Accepts
- Text
- Knowledge cutoff
- Not published
- Weights
- Open
- Licence
- GLM-5.3 License
- Parameters
- 744B total · 40B active
- Reasoning control
- Always on · effort low to max
- API model ID
glm-5.3
Pricing
- Input / 1M tokens
- $1.4
- Cached input / 1M
- $0.26
- Output / 1M tokens
- $4.4
- Blended (3:1)
- $2.15
- Measured cost / task
- $2.01
Worth knowing
- The weights use a custom licence: MIT-style, plus a security-review requirement for very large hosting providers.
Official sources
Closest alternatives
The models nearest to this one on overall score.
23 benchmarks tracked. See methodology and sources.