vs
0:9
overall capability wins (0 ties across 9 scored). GLM-5.3 leads overall.
- Reasoning0:5
- Coding0:1
- Arena0:1
- Specs0:2
Verdict
GLM-5.3 leads coding (widest gap: +25.3 on SciCode); GLM-5.3 ranks higher on LMArena (+117 Elo); GLM-5.3 offers ~5.0x larger context (1M vs 200K).
- GLM-5.3 leads coding (widest gap: +25.3 on SciCode)
- GLM-5.3 ranks higher on LMArena (+117 Elo)
- GLM-5.3 offers ~5.0x larger context (1M vs 200K)
- GLM-5.3 allows ~7.8x more max output (128K vs 16K)
- GLM-5.3 ships open weights (GLM-5.3 License)
Cost for 1M in + 250K out
Illustrative chat workload at primary-provider list prices.
- GLM-4.7 Flash
- $0.16
- GLM-5.3
- -
Price per 1M tokens
Full comparison
Organization
Zhipu AI
Zhipu AI
Family
GLM
GLM
License
Proprietary
GLM-5.3 License
Open weights
No
Yes
Release
Dec 10, 2025
Aug 14, 2026
Knowledge cutoff
-
-
API / provider
Zhipu AI
No primary API price listed
Modalities
text, image → text
text → text
Specs
Context window
200K
1M
Max output
16K
128K
Parameters
—
744B (40B act.)
Pricing
Input $/1M
$0.06
—
Output $/1M
$0.40
—
Blended $/1M (3∶1)
$0.15
-
Speed
tok/s
130
-
TTFT (s)
0.22
-
Reasoning
MMLU-Pro
—
—
GPQA Diamond
58.1
91.7
Humanity's Last Exam
7.6
42.3
AIME 2025
—
—
MATH-500
—
—
Humanity's Last Exam (with tools)
—
—
AA-Omniscience Accuracy
16.2
33.9
AA-LCR v1.1
41.7
79.7
CritPt
0.3
19.1
MMMU-Pro
—
—
IFBench
60.8
—
Chartography
—
—
Chartography (With Tools)
—
—
Coding
SWE-bench Verified
—
—
SWE-bench Pro
—
—
SWE-bench Multilingual
—
—
LiveCodeBench
—
—
Terminal-Bench 2.1
—
83.9
Aider Polyglot
—
—
Terminal-Bench 3
—
28.3
BigCodeBench
—
—
SciCode
33.7
59.0
CursorBench
—
—
SWE-Rebench
30.1
—
NL2Repo-Bench
—
58.0
DeepSWE
—
66.9
WebDev Arena
—
1,614
Terminal-Bench 4.0
—
41.8
CursorBench 4.0
—
—
FrontierCode v1.1 (Main)
—
—
Arena
LMArena Elo
1,366
1,483
| Metric | GLM-4.7 Flash | GLM-5.3 | Delta |
|---|---|---|---|
| Identity | |||
| Organization | Zhipu AI | Zhipu AI | — |
| Family | GLM | GLM | — |
| License | Proprietary | GLM-5.3 License | — |
| Open weights | No | Yes | — |
| Release | Dec 10, 2025 | Aug 14, 2026 | — |
| Knowledge cutoff | - | - | — |
| API / provider | Zhipu AI | No primary API price listed | — |
| Modalities | text, image → text | text → text | — |
| Specs | |||
| Context window | 200K | 1M | B +800K |
| Max output | 16K | 128K | B +112K |
| Parameters | — | 744B (40B act.) | — |
| Pricing | |||
| Input $/1M | $0.06 | — | — |
| Output $/1M | $0.40 | — | — |
| Blended $/1M (3∶1) | $0.15 | - | — |
| Speed | |||
| tok/s | 130 | - | — |
| TTFT (s) | 0.22 | - | — |
| Reasoning | |||
| MMLU-Pro | — | — | — |
| GPQA Diamond | 58.1 | 91.7 | B +33.6 pts |
| Humanity's Last Exam | 7.6 | 42.3 | B +34.7 pts |
| AIME 2025 | — | — | — |
| MATH-500 | — | — | — |
| Humanity's Last Exam (with tools) | — | — | — |
| AA-Omniscience Accuracy | 16.2 | 33.9 | B +17.7 pts |
| AA-LCR v1.1 | 41.7 | 79.7 | B +38.0 pts |
| CritPt | 0.3 | 19.1 | B +18.8 pts |
| MMMU-Pro | — | — | — |
| IFBench | 60.8 | — | — |
| Chartography | — | — | — |
| Chartography (With Tools) | — | — | — |
| Coding | |||
| SWE-bench Verified | — | — | — |
| SWE-bench Pro | — | — | — |
| SWE-bench Multilingual | — | — | — |
| LiveCodeBench | — | — | — |
| Terminal-Bench 2.1 | — | 83.9 | — |
| Aider Polyglot | — | — | — |
| Terminal-Bench 3 | — | 28.3 | — |
| BigCodeBench | — | — | — |
| SciCode | 33.7 | 59.0 | B +25.3 pts |
| CursorBench | — | — | — |
| SWE-Rebench | 30.1 | — | — |
| NL2Repo-Bench | — | 58.0 | — |
| DeepSWE | — | 66.9 | — |
| WebDev Arena | — | 1,614 | — |
| Terminal-Bench 4.0 | — | 41.8 | — |
| CursorBench 4.0 | — | — | — |
| FrontierCode v1.1 (Main) | — | — | — |
| Arena | |||
| LMArena Elo | 1,366 | 1,483 | B +117 Elo |
GLM-4.7 Flash
- Context
- 200K / 16K out
- Parameters
- —
- Price
- $0.06 / $0.40
- Speed
- 130 tok/s · 0.22s TTFT
- Modalities
- text, image → text
- License
- Proprietary
Fast low-cost GLM-4.7 API tier for Chinese/English high-volume chat and light agents.
GLM-5.3
- Context
- 1M / 128K out
- Parameters
- 744B (40B act.)
- Price
- —
- Speed
- —
- Modalities
- text → text
- License
- GLM-5.3 License
Zhipu's Aug 14, 2026 flagship — GLM-5.2 base re-post-trained for complex software engineering, terminal, and real-world agent tasks, with SOTA open-weights coding and emergent cybersecurity skills (CyberGym 84.5). Open weights (744B MoE / 40B active) are on Hugging Face under the GLM-5.3 license; thinking effort is low/high/max (default max).
Benchmark charts
Winner bars are emphasized. Per-benchmark deltas sit above each chart.