Model
Model explorer

GLM-4-32B-0414

OPEN
Zhipu AI · GLM-4-0414 family · released Apr 14, 2025

MIT-licensed 32B base pretrained on 15T tokens (incl. reasoning-heavy synthetic data); competitive with GPT-4o/DeepSeek-V3 on IFEval, BFCL-v3 function calling and TAU-Bench agent tasks at a fraction of the size. Separate 128K-context variant also released.

ReasoningCodingVisionFunction callingTool useAgentic
1393.6
Elo · rank #137
Parameters
32B
Active params
32B (dense)
Context
32K tokens
Architecture
Dense transformer, 32B, pretrained on 15T tokens
License
MIT
Languages
2+
API price (in/out)
No hosted API
Modalities
text
Benchmark results
Bar shows position within the tracked field; marker = field best
BFCL v3Agents69.6%#10
best: Hunyuan-A13B · 78.3%
IFEvalReasoning87.6%#37
best: Gemma 4 26B A4B · 98.5%
SWE-bench VerifiedCoding33.8%#108
best: Claude Fable 5 · 95.0%
tau-benchAgents68.7%#9
best: Claude Opus 4.1 · 82.4%
Run it locally
VRAM @ Q4
20 GB
VRAM @ FP16
64 GB
Fits on (Q4)
RTX 3090 24GBRTX 4090 24GBRTX 5090 32GBM4 Pro 48GBM3 Max 128GBM3 Ultra 512GBA100 80GBH100 80GBH200 141GBB200 192GB
Third-party calculators estimate ~30 tok/s at Q4 on RTX 4090 (extrapolated, not a measured llama.cpp benchmark).
Quantizations
GGUF Q4 · GGUF Q8 · MLX · GPTQ · FP8
GLM-4-0414 family
Elo progression across releases
GLM-4-32B-0414Apr 20251393.6
Resources
HFzai-org/GLM-4-32B-0414GHzai-org/GLM-4APIAvailable via Zhipu AI API
API price weights · each benchmark row carries its own source badge (see methodology)