GLM 4.6
Z.ai · released Sep 30, 2025
Compared with GLM-4.5, this generation brings several key improvements: Longer context window: The context window has been expanded from 128K to 200K tokens, enabling the model to handle more complex...
Specification
- Context window
- 205K
- Max output
- 131K
- Knowledge cutoff
- Mar 31, 2025
- Parameters
- 356.8B
- Licence
- mit
- Serving providers
- 5
- Moderated
- No
- Uptime
- 100.0%
Intelligence
23.4
65th percentile
Coding
—
Coding Index
Agentic
18.6
Agentic Index
Output speed
56 t/s
Median across providers
Latency
451ms
Time to first token
Cost per task
$0.18
Estimated
Benchmarks
Where the score comes from
The Intelligence Index is a composite. These are the underlying evaluations this model was actually measured on.
Evaluation scores
Percentage correct · higher is better
- MMLU-Pro78.4%
- τ²-bench (Telecom)76.9%
- GPQA Diamond63.2%
- LiveCodeBench56.1%
- AIME 202544.3%
- IFBench36.7%
- SciCode33.1%
- Terminal-Bench Hard28.8%
- AA-LCR (long context)28.3%
- Humanity's Last Exam5.5%
An evaluation missing from this list was not run for this model — it is not a zero.
View as table
| Evaluation | Score |
|---|---|
| MMLU-Pro | 78.4% |
| τ²-bench (Telecom) | 76.9% |
| GPQA Diamond | 63.2% |
| LiveCodeBench | 56.1% |
| AIME 2025 | 44.3% |
| IFBench | 36.7% |
| SciCode | 33.1% |
| Terminal-Bench Hard | 28.8% |
| AA-LCR (long context) | 28.3% |
| Humanity's Last Exam | 5.5% |
Against its peers
Intelligence Index · this model highlighted, nearest peers in grey
- Qwen3 MaxQwen24.5
- Ling 3.0 Tiny (free)InclusionAI24.3
- Claude Haiku 4.5Anthropic24.1
- gpt-oss-120bOpenAI24.1
- Kimi K2 0905Moonshot AI24.0
- o1OpenAI23.9
- GLM 4.6Z.ai23.4
- GLM 4.7 FlashZ.ai23.3
Peers are the models sitting closest on the Intelligence Index — the set you would realistically choose between.
View as table
| Model | Intelligence |
|---|---|
| Qwen3 Max | 24.5 |
| Ling 3.0 Tiny (free) | 24.3 |
| Claude Haiku 4.5 | 24.1 |
| gpt-oss-120b | 24.1 |
| Kimi K2 0905 | 24.0 |
| o1 | 23.9 |
| GLM 4.6 | 23.4 |
| GLM 4.7 Flash | 23.3 |
Percentile among all indexed models
Pricing
What it costs to run
List prices per million tokens, plus what one representative task works out to.
List price
- Input / 1M tokens
- $0.5
- Output / 1M tokens
- $2
- Cached input / 1M
- $0.1
- Blended 3:1
- $0.875
One task, estimated
$0.18
- Input tokens
- 50,000
- Output tokens
- 80,000
- Profile
- Reasoning
Estimated from list pricing: 50K input tokens plus 80K output tokens for reasoning models (25K for non-reasoning).
Serving providers
Speed and latency figures are medians across these providers, so a widely-served model reports a blend rather than any single endpoint.
Arena
Head-to-head generation quality
Elo from pairwise judgements, broken out by the kind of thing the model was asked to build.
- Overall Elo
- 1,197
- Win rate
- 54.4%
- Strongest at
- Websites
- Tournaments
- 12,417
Elo by category
Dot position on a 1,050–1,200 scale · Elo has no meaningful zero
- 1,197
- 1,193
- 1,191
- 1,190
- 1,183
- 1,156
- 1,155
- 1,072
Agent categories (full-stack apps, mobile apps) are only scored for models tested in agent harnesses.
View as table
| Category | Elo | Win rate |
|---|---|---|
| Websites | 1197 | 54.4% |
| UI components | 1193 | 53.9% |
| Game development | 1191 | 54.7% |
| Data visualisation | 1190 | 52.6% |
| 3D scenes | 1183 | 54.0% |
| SVG | 1156 | 52.1% |
| Mobile apps | 1155 | 46.8% |
| Full-stack apps | 1072 | 42.3% |