G
GLM-5.3
First among open models on Terminal-Bench 4.0 (38.9%) — the pick for long-horizon agent workflows
Intelligence v5
32.0
Leaderboard rank
#14
Input / output
$0.60/$2.20
Blended 3:1
$1.00
Output speed
77 tok/s
Value
32 pts/$
Capability profile Normalised to the field maximum
Specifications
Vendor
Z.ai
Released
2026-08
Weights
Open (GLM-5.3 License)
Context
1,000,000 tokens
Max output
Undisclosed
Parameters
753B total / 18B active
Input price
$0.60 / 1M
Output price
$2.20 / 1M
Cache hit
n/a
Output speed
77 tok/s
Intelligence
32.0（v5）
Benchmark scores
How it is measured →Ranked 14 · bars are scaled to the field maximum for that benchmark.
| Benchmarks | Score | Vs. best in field | Gap |
|---|---|---|---|
| AA Intelligence Index v5 Intelligence index | 32.0 | 54% | −27.4 pts |
| AA Intelligence Index v4.3.2 Intelligence index (legacy) | 44.8 | 78% | −12.8 pts |
| GPQA Diamond Graduate-level science reasoning | 90.9 | 94% | −5.5 pts |
| Terminal-Bench 4.0 Terminal-agent tasks | 38.9 | 60% | −26.3 pts |
| LiveBench Design | — | — | not indexed |
| Vals Index | — | — | not indexed |
Assessment
753B total with 18B active and 1M context. Its Terminal-Bench 4.0 score of 38.9% is the highest of any open model, double DeepSeek V4.1 Flash's 19.7%. But at $2.01 per index task it costs 7.7× DeepSeek, because it burns more reasoning tokens.
At the same price
The four models closest on blended price, side by side.
| Model | v5 | Blended | tok/s | Score gap |
|---|---|---|---|---|
| MiMo-V2.6 Pro Open weightsXiaomi | 30.0 | $0.61 | 81 | -2.0 |
| DeepSeek V4.1 Flash Open weightsDeepSeek | 32.5 | $0.53 | 148 | +0.5 |
| MiniMax M3.1 FlashMiniMax | 29.0 | $0.26 | 118 | -3.0 |
| Claude Haiku 5.5Anthropic | 36.2 | $0.20 | 235 | +4.2 |
Compare side by side
Line it up against other models on price, speed and every benchmark.
The comparison tool lines up to 4 models side by side with switchable normalisation — built for purchasing decisions.
Open the comparison tool → Back to leaderboard